Pm Skill CreatorSAFE

Product Management skills framework built on battle-tested methods for Claude Code, Cowork, Codex, and AI agents.

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: pm-skill-creator
argument-hint: "[idea or raw content]"
description: Design a new PM skill through guided conversation. Use when you have raw content or an idea and want to shape it into a compliant skill.
intent: >-
  Walk through the full skill design process interactively — from raw idea or content to a structured, repo-compliant SKILL.md draft. Asks adaptive questions to determine skill type, scope, structure, and content, then generates a ready-to-validate draft.
type: interactive
theme: meta-authoring
best_for:
  - "Turning a rough idea or framework into a structured PM skill"
  - "Deciding whether raw content should be a component, interactive, or workflow skill"
  - "Getting from blank page to a complete SKILL.md draft through guided conversation"
scenarios:
  - "I have notes from a workshop and want to turn them into a skill"
  - "Help me create a new skill — I know the topic but not the structure"
  - "I have a PM framework I want to formalize as a repo skill"
estimated_time: "10-15 min"
---

## Purpose

Walk through the full skill design process interactively — from raw idea or content to a structured, repo-compliant SKILL.md draft. Asks adaptive questions to determine skill type, scope, structure, and content, then generates a ready-to-validate draft.

This skill is the conversational complement to `skill-authoring-workflow`. That skill defines the process and validation gates. This one sits with you and figures out what to build before you build it.

## Input

Bring the seed for the skill you want to design — half-formed is fine; shaping it is this skill's job:
- **Works best with:** the idea, framework, or raw content (notes, prompt sequence, article) the skill should capture
- **Also useful:** the skill type you suspect it is, who it's for, and any existing skills it should relate to

If you supply this inline with your request (e.g., "help me design a skill from my pricing-tiers framework"), the conversation uses that as the answer to Q1 and moves on — it won't re-ask. With no input, it opens by asking what you want the skill to help someone do.

Example: `Use pm-skill-creator: I have a decision tree for choosing north-star metrics — help me shape it into a skill.`

## Key Concepts

### When to Use This vs. Other Authoring Tools

| Tool | Best When |
|------|-----------|
| **This skill (skill-creator)** | You have an idea or raw content and need help shaping it into the right structure through conversation |
| `skill-authoring-workflow` | You already know what to build and need the process checklist and validation steps |
| `scripts/build-a-skill.sh` | You know the structure and want a terminal wizard to collect sections |
| `scripts/add-a-skill.sh` | You have a source document and want AI-assisted generation end-to-end |

### The Three Skill Types (Decision Criteria)

- **Component**: One artifact or template. Self-contained. Gets referenced by other skills. Ask: "Is this a thing someone creates?"
- **Interactive**: Guided conversation with adaptive questions and enumerated recommendations. Ask: "Does this require back-and-forth to be useful?"
- **Workflow**: Multi-phase orchestration referencing other skills. Ask: "Does this coordinate multiple activities across steps?"

### Skill Anatomy (Non-Negotiable Sections)

Every skill requires these sections in order:
1. **Purpose** — What it does + when to use it (outcome-focused)
2. **Input** — What the user can bring (subject, optional context), an example invocation, and the rule that inline input counts as answers already given. Must read as an invitation: zero input is fine, the guided flow covers it. Plain language only — no runtime template syntax like `$ARGUMENTS`
3. **Key Concepts** — Frameworks, definitions, mental models
4. **Application** — Step-by-step instructions an agent can follow
5. **Examples** — At least one concrete, specific example
6. **Common Pitfalls** — Named failure modes with consequences and fixes
7. **References** — Related skills, external frameworks, source material

### Metadata Constraints

- `name`: lowercase kebab-case, ≤ 64 characters
- `description`: ≤ 200 characters, must include trigger cue ("Use when...")
- `intent`: longer repo-facing summary (no character limit)
- `type`: one of `component`, `interactive`, `workflow`

### Facilitation Source of Truth

Use [`workshop-facilitation`](../workshop-facilitation/SKILL.md) as the default interaction protocol for this skill.

It defines:
- Session heads-up + entry mode (Guided, Context dump, Best guess)
- One-question turns with plain-language prompts
- Progress labels (e.g., Context Q1/5)
- Interruption handling and pause/resume behavior
- Numbered recommendations at decision points
- Quick-select numbered response options (include `Other (specify)` when useful)

This file defines the domain-specific content. If there is a conflict, follow this file's domain logic.

## Application

This interactive skill asks **up to 5 adaptive questions**, then delivers a **complete SKILL.md draft** with frontmatter, all required sections, and repo-compliant structure.

---

### Step 0: Session Start

**Agent says:**

"I'll help you design a new PM skill from scratch. This takes about 10-15 minutes and up to 5 questions. How do you want to start?

1. **Guided** — I'll ask questions one at a time and build the skill from your answers (recommended)
2. **Context dump** — Paste your raw content, notes, or framework and I'll propose a skill structure
3. **Best guess** — Tell me just the topic and I'll draft a skill you can refine"

---

### Question 1: What's the Raw Material? (Q1/5)

**Agent asks:**

"What are we turning into a skill? Give me whatever you have."

1. **A framework or mental model** — A structured way of thinking about a PM problem (e.g., prioritization matrix, decision tree)
2. **A template or artifact** — A deliverable PMs create (e.g., PRD section, positioning statement, epic format)
3. **A process or workflow** — A multi-step method 

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (0)

No findings outside the package's declared scope.

Gates applied: no_behavioural_pass.