skill-authoring-workflow
Turn raw PM content into a compliant, publish-ready skill. Use when creating or updating a repo skill without breaking standards.
Security Assessment
About skill-authoring-workflow
The skill-authoring-workflow is a multi-phase orchestration skill for turning raw product-management material — rough notes, workshop transcripts, frameworks, or half-baked prompt dumps — into compliant, publish-ready `skills/<skill-name>/SKILL.md` assets that pass the repository's validation standards. It exists to solve the problem of ad-hoc, inconsistent skill authoring, replacing 'looks good to me' guesswork with a repeatable process that keeps repo standards intact when creating or updating a skill.
The workflow runs six phases: preflight (searching for overlapping skills and deciding the skill type — component, interactive, or workflow), draft generation, tightening, hard validation, repo-doc integration, and optional packaging. It leans on repo-native tooling rather than inventing custom processes: shell scripts such as `find-a-skill.sh`, `add-a-skill.sh`, `build-a-skill.sh`, and `test-a-skill.sh`, plus Python metadata and trigger checkers. A strict Definition of Done enforces valid frontmatter, compliant section ordering, metadata character limits, resolvable cross-references, and updated README catalog counts. It uses a companion `workshop-facilitation` skill as its interactive facilitation protocol for guided, one-question-at-a-time sessions.
The target users are contributors to Dean Peters' product-manager-skills repository who want to author or revise skills without breaking its conventions. It fits PMs and skill maintainers converting workshop content into shareable assets, and anyone who needs the full authoring-and-validation loop run before committing. It is tightly coupled to one specific repository's scripts and standards, so its value outside that repo is limited to the general workflow discipline it models.
FAQ
What do I need to provide to use this workflow?
You bring the raw source material (notes, transcript, framework, or prompt sequence) or the existing skill you want to update. Optionally you can supply the intended skill type, target audience, and a naming preference. Arriving with partial or zero input is fine — the workflow will ask what you want to turn into a skill if you provide nothing.
What are the three skill types it supports?
Component (a single artifact or template), Interactive (three to five adaptive questions with numbered options), and Workflow (multi-phase orchestration). Phase 1 helps you decide which fits before drafting.
When is a skill considered done?
Only when frontmatter is valid, section order is compliant, metadata limits are respected (name under 64 chars, description under 200), the description states both what and when, cross-references resolve, and README catalog counts are updated. There are no exceptions to the Definition of Done.
Does it work outside the product-manager-skills repository?
The workflow depends on repo-native scripts like find-a-skill.sh, add-a-skill.sh, and check-skill-metadata.py, so it is designed for that specific repo. The general authoring discipline transfers, but the exact commands assume that repository's structure.
Install skill-authoring-workflow
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
deanpeters/product-manager-skills