Skill Authoring WorkflowSAFE

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: skill-authoring-workflow
argument-hint: "[source content or skill to update]"
description: Turn raw PM content into a compliant, publish-ready skill. Use when creating or updating a repo skill without breaking standards.
intent: >-
  Create or update PM skills without chaos. This workflow turns rough notes, workshop content, or half-baked prompt dumps into compliant `skills/<skill-name>/SKILL.md` assets that actually pass validation and belong in this repo.
type: workflow
best_for:
  - "Creating a new repo skill from notes or source material"
  - "Updating an existing skill while keeping standards intact"
  - "Running the full authoring and validation workflow before commit"
scenarios:
  - "Help me turn these workshop notes into a new PM skill"
  - "I need to update an existing skill without breaking the repo standards"
  - "What workflow should I use to author a new skill in this repo?"
theme: meta-authoring
estimated_time: "45-90 min"
---

## Purpose

Create or update PM skills without chaos. This workflow turns rough notes, workshop content, or half-baked prompt dumps into compliant `skills/<skill-name>/SKILL.md` assets that actually pass validation and belong in this repo.

Use it when you want to ship a new skill without "looks good to me" roulette.

## Input

Bring the raw material and the intent — rough is fine; the workflow exists to get it the rest of the way:
- **Works best with:** the source content (notes, transcript, framework, prompt sequence) or the existing skill you want to update
- **Also useful:** the intended skill type (component/interactive/workflow), target audience, and any naming preference

If you supply this inline with your request (e.g., "turn `research/pricing-workshop-notes.md` into an interactive advisor"), the workflow starts at Phase 1 with that context — it won't re-ask for what you already gave. If you provide nothing, it opens by asking what content you want to turn into a skill and offers the entry modes from the facilitation protocol.

Example: `Use skill-authoring-workflow: convert research/pricing-workshop-notes.md into an interactive pricing advisor.`

## Key Concepts

### Dogfood First

Use repo-native tools and standards before inventing a custom process:
- `scripts/find-a-skill.sh`
- `scripts/add-a-skill.sh`
- `scripts/build-a-skill.sh`
- `scripts/test-a-skill.sh`
- `scripts/check-skill-metadata.py`

### Pick the Right Creation Path

- **Guided wizard (`build-a-skill.sh`)**: Best when you have an idea but not final prose.
- **Content-first generator (`add-a-skill.sh`)**: Best when you already have source content.
- **Manual edit + validate**: Best for tightening an existing skill.

### Definition of Done (No Exceptions)

A skill is done only when:
1. Frontmatter is valid (`name`, `description`, `intent`, `type`)
2. Section order is compliant (Purpose, Input, Key Concepts, Application, Examples, Common Pitfalls, References)
3. Metadata limits are respected (`name` <= 64 chars, `description` <= 200 chars)
4. Description says both what the skill does and when to use it
5. The Input section says what the user can bring, shows an example invocation, tells the agent to use inline input instead of re-asking, and makes clear that arriving with partial or zero input is fine — in plain language, never runtime template syntax like `$ARGUMENTS` (rationale: CONTRIBUTING.md, "Why We Don't Use `$ARGUMENTS`")
6. Intent carries the fuller repo-facing summary without replacing the trigger-oriented description
7. Cross-references resolve
8. README catalog counts and tables are updated (if adding/removing skills)

### Facilitation Source of Truth

When running this workflow as a guided conversation, use [`workshop-facilitation`](../workshop-facilitation/SKILL.md) as the interaction protocol.

It defines:
- session heads-up + entry mode (Guided, Context dump, Best guess)
- one-question turns with plain-language prompts
- progress labels (for example, Context Qx/8 and Scoring Qx/5)
- interruption handling and pause/resume behavior
- numbered recommendations at decision points
- quick-select numbered response options for regular questions (include `Other (specify)` when useful)

This file defines the workflow sequence and domain-specific outputs. If there is a conflict, follow this file's workflow logic.

## Application

### Phase 1: Preflight (Avoid Duplicate Work)

1. Search for overlapping skills:

```bash
./scripts/find-a-skill.sh --keyword "<topic>"
```

2. Decide type:
- **Component**: one artifact/template
- **Interactive**: 3-5 adaptive questions + numbered options
- **Workflow**: multi-phase orchestration

### Phase 2: Generate Draft

If you have source material:

```bash
./scripts/add-a-skill.sh research/your-framework.md
```

If you want guided prompts:

```bash
./scripts/build-a-skill.sh
```

### Phase 3: Tighten the Skill

Manually review for:
- Clear "when to use" guidance
- One concrete example — optimally two, from different business domains (one SaaS, one industrial/non-SaaS), so the framework visibly generalizes; reuse the repo's fictional universes (Fieldlight/Wrenchline for SaaS, Helix/Northfield/Corvid for industrial) and suffix the second file by domain (`sample-industrial.md`)
- A `template.md` when the skill produces an artifact — the output schema as a copy/paste fill-in with quality checks
- One explicit anti-pattern
- No filler or vague consultant-speak

### Phase 4: Validate Hard

Run strict checks before thinking about commit:

```bash
./scripts/test-a-skill.sh --skill <skill-name> --smoke
python3 scripts/check-skill-metadata.py skills/<skill-name>/SKILL.md
python3 scripts/check-skill-triggers.py skills/<skill-name>/SKILL.md --show-cases
```

### Phase 5: Integrate with Repo Docs

If this is a new skill:
1. Add it to the correct README category table
2. Update skill totals and category counts
3. Verify link paths resolve

### Phase 6: Optional Packaging

If targeting Claude custom skill upload:

```bash
./scripts/zip-a-skill.sh

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.