Ready-to-Download AI Agent Skills

Try a request

• Review these fictional partner profiles and example referral links.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Qualify partners and verify referral links without mixing commercial terms between clients.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Approved partner criteria
• —Program terms
• —Product destinations
• —Attribution requirements

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md
• references/partner-pipeline.md

Try a request

• Prepare an onboarding gap list and first-workflow runbook for a fictional agency client.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Turn a new client engagement into verified workspace setup, an owned first workflow and a maintainable runbook.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Engagement scope and success criteria
• —Client workspace identity and approved context
• —Available systems and owners
• —Existing setup and handoff artifacts

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md
• references/runbook.md

Try a request

• Create a measurement plan for a fictional signup funnel.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Define and verify events, conversions and campaign attribution for a measurable funnel.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Business funnel
• —Existing event inventory
• —Consent requirements
• —Campaign naming rules

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md

Try a request

• Design a pagination adapter from this synthetic API specification.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Design an API integration from official contracts and test its data and error handling.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Provider documentation
• —Approved capability
• —Input/output contract
• —Example non-sensitive responses

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md

Try a request

• Diagnose this synthetic workflow’s duplicate-record behavior.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Design and diagnose reliable workflows with explicit data ownership and failure handling.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Workflow export
• —Input and output contracts
• —Field ownership
• —Retry and volume constraints

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md

Try a request

• Prepare a production brief for an authorized fictional avatar.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Prepare and verify avatar videos using approved likenesses, scripts and provider access.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Authorized avatar and voice
• —Exact script
• —Aspect ratio and language
• —Approved video provider

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md

Try a request

• Create a management report from a fictional table and verify its totals.

Host requirements

• Authorized local file tools

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Produce and verify Word, PDF, presentation and spreadsheet deliverables using available host tools.

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Content and source data
• —Required file format
• —Approved template
• —Audience and delivery requirements

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md

Try a request

• Find potential buyers from these fictional company and hiring records, with evidence for each qualification.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Find and qualify prospective companies using dated evidence, ICP fit and duplicate-aware research.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Workspace ICP and exclusions
• —Search markets and evidence window
• —Existing prospect inventory
• —Qualification rubric

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md
• references/qualification.md

Try a request

• Use Content Production Pipeline for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, required access and output in the active Workspace before client use.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Takes raw source material through intake, editing, derivative generation and per-channel task creation, tracking each asset's state.

Services it may use

What it reads from Workspace Context

• —Organization and client identity
• —Goals and KPIs
• —Approval policy and named approvers
• —The connected-system inventory for this workspace
• —Brand voice and reference assets

What you still supply

• —The source material and its intake location
• —The channel matrix and per-channel requirements

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —Derivatives are produced as drafts; publication is a separate approved step

Known limitations

• —Never modifies the original source asset
• —Does not publish — hand off to a publishing Skill
• —Labelled Limited guided pilot

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml

Try a request

• Use Creative Production and Briefing for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, required access and output in the active Workspace before client use.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Turns a performance finding into a production-ready creative brief, grounded in the client's real product and official brand marks.

Services it may use

What it reads from Workspace Context

• —Organization and client identity
• —Goals and KPIs
• —Approval policy and named approvers
• —The connected-system inventory for this workspace
• —Brand assets, reference material and claims policy

What you still supply

• —The performance finding that triggered the brief
• —The claims policy and any legal constraints

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —The brief and any candidate visuals are routed for approval before use

Known limitations

• —Does not buy media or edit campaigns
• —Will not approximate a logo or product it cannot retrieve authoritatively
• —Labelled Limited guided pilot

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml

Try a request

• Design a CRM for a fictional service company using these sample tables.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Design a maintainable CRM schema and migration plan from the business process.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Current schema export
• —Record lifecycle
• —Ownership rules
• —Required reports

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md

Try a request

• Use CRM Data Hygiene for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, required access and output in the active Workspace before client use.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Audits CRM records for missing, contradictory or unusable data and reports exactly what a human must fix.

Services it may use

What it reads from Workspace Context

• —Organization and client identity
• —Goals and KPIs
• —Approval policy and named approvers
• —The connected-system inventory for this workspace

What you still supply

• —The status vocabulary and what each status actually means — the real values, not the labels

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —None — this Skill reports and changes nothing

Known limitations

• —Reports only; it never edits a record
• —Cannot infer a status vocabulary that has not been documented

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml

Try a request

• Use CRM Revenue Recovery for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, required access and output in the active Workspace before client use.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Finds revenue already in the CRM that is quietly going cold and nudges the responsible person with the evidence attached.

Services it may use

What it reads from Workspace Context

• —Organization and client identity
• —Goals and KPIs
• —Approval policy and named approvers
• —The connected-system inventory for this workspace
• —Approval policy and who owns which pipeline

What you still supply

• —The written recovery rules — what counts as overdue or stalled
• —The owner roster

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —Any outbound nudge is shown in full and sent only after approval

Known limitations

• —Does not change deal stages on its own
• —Does not contact the end customer

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml

Try a request

• Use Cross-Channel Performance Reporting for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, required access and output in the active Workspace before client use.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Pulls performance across every connected channel onto one measurement basis, explains what changed and why, and prepares a client-ready report for release.

Services it may use

What it reads from Workspace Context

• —Organization and client identity
• —Goals and KPIs
• —Approval policy and named approvers
• —The connected-system inventory for this workspace
• —Contact roster and email or output formatting rules
• —Performance marketing goals

What you still supply

• —The reporting period
• —Which channels are in scope
• —The recipient and delivery format

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —The report is prepared as a draft; a named approver releases it. The Skill never sends to a client directly

Known limitations

• —Does not reconcile against an accounting system
• —Cannot invent a measurement basis the workspace has not declared

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml

Try a request

• Use Custom API Operations for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, required access and output in the active Workspace before client use.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

A governed template for reaching any approved provider API when no purpose-built workflow exists — preflight, safe read, bounded write, read-back.

Services it may use

What it reads from Workspace Context

• —Organization and client identity
• —Goals and KPIs
• —Approval policy and named approvers
• —The connected-system inventory for this workspace

What you still supply

• —The provider's official API documentation
• —The specific operation and its intended scope

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —Runtime and Workspace policy govern writes and verification. Destructive operations are outside this capability’s scope.

Known limitations

• —Never constructs its own authorization header or uses a found key
• —Will not guess an endpoint — it reads the official documentation or stops

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml

Try a request

• Use Data Sync and Workflow Orchestration for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, required access and output in the active Workspace before client use.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Reconciles records between connected systems with field mappings, duplicate checks, conflict rules and a report for each run.

Services it may use

What it reads from Workspace Context

• —Organization and client identity
• —Goals and KPIs
• —Approval policy and named approvers
• —The connected-system inventory for this workspace

What you still supply

• —A written field mapping
• —The match key
• —The conflict rule when both sides changed

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —The first run of any new mapping is proposed and approved. Destructive reconciliation is separately gated

Known limitations

• —Will not run without a written match key and conflict rule
• —Does not design the destination schema

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml

Try a request

• Audit these synthetic headers and DNS records for alignment problems.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Diagnose email authentication and reputation problems with a prioritized remediation plan.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Approved sending domains
• —DNS observations
• —Message headers
• —Delivery and bounce exports

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md

Try a request

• Use Email Follow-Up for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, required access and output in the active Workspace before client use.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Finds inbound messages that were never answered, drafts a reply in the client's voice, and creates a draft for a named approver rather than sending.

Services it may use

What it reads from Workspace Context

• —Organization and client identity
• —Goals and KPIs
• —Approval policy and named approvers
• —The connected-system inventory for this workspace
• —Contact roster, communication rules and email formatting

What you still supply

• —The mailbox and time window in scope
• —Tone and signature rules if not already in Workspace Context

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —Creates drafts only. A human sends. The Skill never sends mail itself

Known limitations

• —Never sends
• —Does not manage deliverability, DNS or warmup

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml

Try a request

• Diagnose these synthetic authentication results and prepare a sending-domain onboarding checklist.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Audit sending domains, diagnose authentication and reputation, and prepare verified mailbox onboarding.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Sending domain and mailbox inventory
• —DNS/provider access
• —Current authentication records and message headers
• —Approved sending pattern and consent policy

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/diagnostics.md
• references/evaluation.md

Try a request

• Draft a welcome sequence and test its segment rules on synthetic contacts.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Build email segments, templates and lifecycle campaigns from approved audience rules.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Audience consent and suppression rules
• —Campaign objective
• —Approved content
• —Provider capability

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md
• references/provider-planning.md

Try a request

• Use Facebook and Instagram Community Operations for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, required access and output in the active Workspace before client use.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Reads public comments and mentions, classifies each by intent, drafts replies from the client's knowledge base, and posts only what an approver signed off.

Services it may use

What it reads from Workspace Context

• —Organization and client identity
• —Goals and KPIs
• —Approval policy and named approvers
• —The connected-system inventory for this workspace
• —Brand voice and escalation policy

What you still supply

• —The knowledge base replies should be drawn from
• —The escalation rule for complaints and sensitive topics

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —Every reply is approved individually by a named client approver before it is posted

Known limitations

• —Public comments and mentions only — no private inbox or direct messages
• —Does not moderate or delete third-party content without approval
• —Labelled Limited guided pilot

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml

Try a request

• Review this fictional settings screen for keyboard and mobile usability.

Host requirements

• Authorized local file tools

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Improve a product interface’s hierarchy, accessibility and interaction states.

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —UI source or screenshots
• —User task
• —Existing design system
• —Target devices

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md

Try a request

• Prepare a paused search campaign build from this synthetic keyword list.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Turn an approved search strategy into an inspectable campaign build and launch checklist.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Offers and landing pages
• —Target geography and language
• —Approved budget
• —Conversion definitions

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md

Try a request

• Use Google Ads Optimization and Governance for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, required access and output in the active Workspace before client use.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Audits a connected Google Ads account, classifies every search term, and turns waste into an approval package that only a named human can release.

Services it may use

What it reads from Workspace Context

• —Organization and client identity
• —Goals and KPIs
• —Approval policy and named approvers
• —The connected-system inventory for this workspace
• —Performance marketing goals and target cost per acquisition

What you still supply

• —The date range to audit
• —The conversion action that counts as the measurement basis
• —Who the named approver is

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —Every negative keyword, bid, budget and structure change is proposed as a batch and applied only after a named approver signs off
• —Nothing is applied on an unverified account identity

Known limitations

• —Does not create campaigns from scratch
• —Does not change the primary conversion action

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/change-conflicts.md

Try a request

• Use Google Drive and Shared Drive File Operations for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, required access and output in the active Workspace before client use.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Audits Drive structure, finds duplicates and orphans, builds the target hierarchy, and moves, renames or uploads files under approval.

Services it may use

What it reads from Workspace Context

• —Organization and client identity
• —Goals and KPIs
• —Approval policy and named approvers
• —The connected-system inventory for this workspace

What you still supply

• —The agreed folder taxonomy
• —Which drive or folder tree is in scope

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —Every move, rename, copy and upload is proposed and applied only after approval

Known limitations

• —Never permanently deletes — it trashes at most, under approval
• —The organizational logic is untested against a real client tree

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml

Try a request

• Use Google Merchant Center Operations for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, required access and output in the active Workspace before client use.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Reviews a connected product feed for disapprovals, missing attributes and quality gaps, then ranks the fixes by revenue exposure.

Services it may use

What it reads from Workspace Context

• —Organization and client identity
• —Goals and KPIs
• —Approval policy and named approvers
• —The connected-system inventory for this workspace

What you still supply

• —Which product set or feed is in scope
• —The revenue or margin basis used to rank fixes

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —Read and propose only; this revision does not write to the feed.

Known limitations

• —Does not author product data from scratch
• —Does not manage the underlying ecommerce catalog

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml

Try a request

• Produce a weekly GTM scorecard from fictional channel and CRM data; show unknown conversion rates explicitly.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Measure GTM funnel progress, channel yield and next-period priorities from reconciled evidence.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Comparable reporting periods
• —Funnel definitions and attribution rules
• —Verified CRM and channel observations
• —Targets and team capacity

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md
• references/measurement.md

Try a request

• Reconcile a synthetic buyer and partner pipeline and prepare today’s follow-up queue without sending.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Reconcile buyer, partner and follow-up queues into an owned, measurable next-action plan.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —CRM schema and current pipeline
• —Approved outreach channels
• —Owner and contact history
• —Capacity and active GTM priorities

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md
• references/queue.md

Try a request

• Use ICP and Positioning Strategy Builder for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, required access and output in the active Workspace before client use.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Provides strategy methods for turning discovery and verified research into an ICP, differentiation and channel messaging.

Services it may use

What it reads from Workspace Context

• —Organization and client identity
• —Any existing positioning, goals or audience notes already recorded

What you still supply

• —The client's website
• —Discovery call notes or a completed intake brief
• —Research evidence for the relevant conclusions, from existing sources or authorized research

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —The active Workspace policy governs strategy approval and delivery; unsupported claims remain unresolved.

Known limitations

• —Produces strategy documents; it runs no campaign and changes no external system
• —Conclusions that lack evidence stay pending while supported sections can be completed.
• —Public sources only — no authenticated scraping and no stored browser sessions

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/01-intake-brief.md
• references/02-research-commission.md
• references/03-strategy-document-structure.md
• references/04-strategy-methodology.md
• references/05-delivery-note.md
• references/workspace-config.example.md

Try a request

• Triage these fictional Instagram threads into business replies, support handoffs and personal messages to leave alone.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Triage business messages, preserve conversation context and verify authorized replies without mixing personal conversations.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Approved business account and inbox surfaces
• —Current product facts and reply policy
• —Conversation history and message intent
• —Authorized response or draft scope

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md
• references/messages.md

Try a request

• Audit this fictional landing page for mobile form friction.

Host requirements

• Authorized local file tools

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Find conversion obstacles across a page’s message, evidence, forms and mobile experience.

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Page or screenshots
• —Intended audience
• —Primary conversion
• —Available analytics

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md

Try a request

• Use Lead and Call Quality for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, required access and output in the active Workspace before client use.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Cross-checks inbound lead intent against how each lead was classified, and measures first-response time against the agreed service level.

Services it may use

What it reads from Workspace Context

• —Organization and client identity
• —Goals and KPIs
• —Approval policy and named approvers
• —The connected-system inventory for this workspace

What you still supply

• —Written qualification criteria
• —The stated first-response service level

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —None — this Skill reports and changes nothing

Known limitations

• —Reports only; it never reclassifies a lead
• —Requires call-tracking data to assess response time

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml

Try a request

• Audit our MailerLite forms and explain why repeat subscribers miss the lead magnet.
• Prepare an English version of this nurture flow without activating it.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Audit and operate MailerLite forms, subscribers, campaigns and multilingual nurture automations.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Account and product edition
• —Audience consent and suppression rules
• —Form and automation objective
• —Approved copy, sender and destinations

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/audience.md
• references/automations.md
• references/content.md
• references/evaluation.md
• references/sources.md

Try a request

• Design a test using fictional campaign data and a fixed budget.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Turn a marketing hypothesis into a measurable experiment and decision plan.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Objective and baseline
• —Available audience
• —Budget limits
• —Observation window

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md

Try a request

• Design and test a local MCP integration for a synthetic paginated inventory service.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Design and verify bounded MCP tools and agent integrations from explicit contracts and authorized capabilities.

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Integration goal and host
• —Official protocol and provider documentation
• —Tool input/output contracts
• —Authorized development workspace and synthetic fixtures

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/contracts.md
• references/evaluation.md

Try a request

• Use Meta Ads Optimization and Governance for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, required access and output in the active Workspace before client use.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Reads a connected Meta advertising account, detects creative fatigue and delivery decay, and proposes budget and creative action as an approval package.

Services it may use

What it reads from Workspace Context

• —Organization and client identity
• —Goals and KPIs
• —Approval policy and named approvers
• —The connected-system inventory for this workspace
• —Performance marketing goals and target cost per acquisition

What you still supply

• —The reporting window
• —The revenue or conversion source of truth
• —Who the named approver is

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —Budget, status and creative changes are applied only after a named approver signs off

Known limitations

• —App review and scope approval are per-app and unproven for any new client
• —Does not produce creative assets
• —Labelled Limited guided pilot in both capability registries

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml

Try a request

• Use Multi-Channel Social Publishing for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, required access and output in the active Workspace before client use.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Adapts one approved asset to each channel's native format and schedules it, enforcing per-day limits and pre-publish approval.

Services it may use

What it reads from Workspace Context

• —Organization and client identity
• —Goals and KPIs
• —Approval policy and named approvers
• —The connected-system inventory for this workspace
• —Brand voice and communication rules

What you still supply

• —The approved source asset
• —The channel list and schedule
• —Per-day posting limits

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —Everything is held for pre-publish approval; per-day limits are enforced before scheduling

Known limitations

• —Does not create the source asset
• —Does not manage paid promotion
• —Labelled Limited guided pilot

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml

Try a request

• Use Outbound Prospecting for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, required access and output in the active Workspace before client use.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Decides whether a segment is worth attacking, sources and qualifies prospects against written criteria, deduplicates against the pipeline, and prepares a sequenced campaign for approval.

Services it may use

What it reads from Workspace Context

• —Organization and client identity
• —Goals and KPIs
• —Approval policy and named approvers
• —The connected-system inventory for this workspace
• —ICP, positioning and target geography

What you still supply

• —The ICP and written inclusion criteria
• —A suppression list and a current pipeline export

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —No sequence is activated and no message is sent without explicit approval

Known limitations

• —Does not warm domains or manage sending infrastructure
• —Does not write the final copy without review
• —Labelled Limited guided pilot

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml

Try a request

• Use Outloop Custom API Setup for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, the approved browser path and the resulting configuration in the active Workspace before client use.

Runtime dependency: An approved browser path (Outloop Managed Browser, or Claude in Chrome) plus Outloop Local Bridge access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Name a service and the Skill sets it up in Outloop for you — researches the provider, creates the custom service, defines its operations, configures async and file handling, and takes it through provider verification.

Services it may use

What it reads from Workspace Context

• —The workspace (tenant) identity this client folder is mapped to
• —The approved capability list for this workspace
• —The visible state of the Outloop dashboard it is operating

What you still supply

• —The service you want to connect, and what the agent should be able to do with it
• —The provider credential — typed by you, into Outloop, when Outloop asks for it

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —The credential is entered by a person in Outloop and never handled by the Skill. The Skill pauses, says in one sentence what to enter, and resumes on the visible stored state.
• —Workspace access, destructive actions and data-privacy deletion stay the operator's decisions. Nothing is switched on silently.

Known limitations

• —Never asks for, reads, types or records a credential value — no chat field, no example value, no screenshot of one
• —Configures only through the visible Outloop dashboard in a browser. It uses no terminal step, and the one-time Managed Browser MCP registration is operator setup performed before the Skill runs.
• —An OpenAPI import produces a reviewed draft of operations, never an approved or verified workflow
• —Setup is a set of independent facts — saved, operations configured, credentials complete, workspace access on, provider verified — and the Skill never reports a collapsed "connected" state
• —Worked provider recipes are examples of reusable patterns, not adapters or supported connectors

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• evals/README.md
• evals/eval-01-happy-path.md
• evals/eval-02-half-workflow.md
• evals/eval-03-401-composition.md
• evals/eval-04-security-refusals.md
• references/async-and-ids.md
• references/browser-flow.md
• references/browser-security.md
• references/browser-tools.md
• references/capability-model.md
• references/credential-handoff.md
• references/files-and-media.md
• references/patterns/async-create-poll-result.md
• references/patterns/bearer-oauth.md
• references/patterns/complex-apis.md
• references/patterns/downloads.md
• references/patterns/examples/higgsfield.md
• references/patterns/examples/pandadoc.md
• references/patterns/multi-part-credential.md
• references/patterns/provider-generated-ids.md
• references/patterns/provider-media-urls.md
• references/patterns/rest-api-key.md
• references/patterns/soap-session.md
• references/patterns/uploads.md
• references/provider-research.md
• references/service-routing.md
• references/setup-state-machine.md
• references/troubleshooting.md
• references/verification.md

Try a request

• Prepare an outreach review queue from these synthetic prospects.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Turn a qualified segment into a measurable outreach plan and review queue.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —ICP and exclusions
• —Verified prospects
• —Suppression history
• —Offer and channel constraints

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md

Try a request

• Use Performance Growth Operator for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Refactored to canonical OLLIE Runtime ownership with explicit inputs. Local package and calculation tests are separate from Cowork import and live-service proof.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Runs the monthly performance cycle end to end: real economics, full-funnel reporting, budget scenarios, creative sizing and a briefed studio package.

Services it may use

What it reads from Workspace Context

• —Which business model this client is — ecommerce, lead generation or SaaS
• —The client's own margin, close rate and customer value where the Workspace holds them
• —Approved accounts and properties for this Workspace, and which resources it may touch
• —Who approves client-facing delivery, and the agreed reporting cadence

What you still supply

• —The reporting period
• —The client's business economics, if the Workspace does not already hold them — margin, close rate, customer value

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —Client-facing delivery is approval-gated — the report reaches the client only after a person releases it
• —Budget and scaling decisions are recommendations; a human applies them
• —The internal report is sent automatically once the run completes

Known limitations

• —Requires explicit business inputs and approved source access.
• —Cowork installation and live workspace execution remain unverified.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• assets/client-report-template.html
• references/action-engine.md
• references/approval-model.md
• references/benchmark-priors.md
• references/brief-artifacts-and-drive.md
• references/client-report-html.md
• references/client-studio-direct-loop.md
• references/client-vs-internal-reporting.md
• references/creative-intelligence.md
• references/creative-production.md
• references/data-source-doctrine.md
• references/ecommerce-model.md
• references/financial-fallbacks.md
• references/forecasting.md
• references/lead-generation-model.md
• references/meta-video-intelligence.md
• references/saas-model.md
• references/studio-brief.md
• scripts/forecast_scenarios.py
• scripts/html_report_validator.py
• scripts/leak_scan.py
• scripts/profitability_router.py

Try a request

• Reconcile fictional invitation and message records into a follow-up queue.

Host requirements

• Approved access to the services used for this task

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Reconcile relationship status and prepare relevant professional-network follow-ups.

Services it may use

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Approved prospect list
• —Relationship observations
• —Message history
• —Channel policy

What it can change

Produces drafts for a person to review. Nothing is sent or published by the Skill.

What needs human approval

• —Runtime applies the active Workspace authority and approval rules.

Known limitations

• —New generic capability; Cowork execution and provider actions require separate verification.

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml
• references/evaluation.md
• references/relationship-reconciliation.md

Try a request

• Use Project Board and Approval Operations for this Workspace and return the requested output with supporting evidence.

Host requirements

• Authorized host file tools and the service access needed for this task

Compatibility and proof

Instructions aligned with the OLLIE Operator; package checks only. Verify installed discovery, required access and output in the active Workspace before client use.

Runtime dependency: OLLIE Runtime Pack (current supported candidate).

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

Versioned public source; previous download preserved in release history.

What it solves

Scans a project board for stalled, duplicated, unowned and abandoned work and produces an approval-ready cleanup batch.

Services it may use

What it reads from Workspace Context

• —Organization and client identity
• —Goals and KPIs
• —Approval policy and named approvers
• —The connected-system inventory for this workspace
• —Project management conventions

What you still supply

• —The documented approval convention
• —Which boards or projects are in scope

What it can change

Can change connected-system state within current authorization. Runtime enforces the active Workspace approval rules.

What needs human approval

• —Cleanup is applied as an approved batch, never item by item without review

Known limitations

• —Does not close work on someone's behalf without approval
• —Does not reassign owners unilaterally
• —Labelled Limited guided pilot

How to install it

1. 1. Download the complete ZIP and compare its SHA-256 with this page.
2. 2. In Claude/Cowork, open Customize → Skills → Create skill → Upload a skill, then choose the ZIP and enable it. For the full catalog plugin, use Customize → Plugins instead.
3. 3. In Claude Code, extract the named skill folder into your personal .claude/skills directory, preserving every supporting file.
4. 4. For Ask OLLIE, install and verify the separate Runtime Pack. Confirm the skill is discoverable in a fresh session, then select the intended Outloop Workspace.
5. 5. Request the outcome. Check the actual output and required service access; a successful import alone does not verify external execution.

What is in the package

• SKILL.md
• agents/openai.yaml

Try a request

• Reconcile two fictional runbooks and identify conflicting instructions.

Host requirements

• Authorized local file tools

Compatibility and proof

Generic instructions consume canonical OLLIE Runtime owners. Package validation is separate from host installation and live execution.

Runtime dependency: OLLIE Runtime Pack; authorized host tools and required service access.

• Claude / Cowork: Package structure checked; GUI import and execution not verified.
• Claude Code: Folder format checked; model execution not verified.

New generic instructions informed by a private capability inventory; no private source files or proprietary host libraries are redistributed.

What it solves

Reconcile project documents and procedures while preserving evidence and ownership.

What it reads from Workspace Context

• —Current Workspace identity, approved facts, bindings and permissions

What you still supply

• —Current approved documents
• —New evidence
• —Canonical document owners
• —Requested update scope