Review product documents with AI
Review a product document against its purpose, audience and evidence, then choose targeted revisions before sharing it.
Review product documents with AI after establishing what the product does and who the document needs to help. The pm skill uses existing product context or a short session-only interview before examining evidence, contradictions, boundaries and reader needs.
You can stop after the critique or request targeted fixes. Simulated engineering, quality or customer perspectives offer different readings, rather than actual stakeholder feedback. The resulting verdict remains an aid to review, not a readiness certificate or proof that implementation will succeed.
What it does
- Connect a draft to its product purpose
- Find claims without supporting evidence
- Compare dates, numbers and rules for contradictions
- Identify open questions and their likely owners
- Choose targeted revisions without losing author intent
- Route into brief, story or metrics drafting when needed
Install and use
- Load the complete skill and references through the documented Claude Code or Codex route. Prepare a readable draft and its purpose; use the three-question session-only path if persistent context is absent.
- Ask: “Review this search-change brief for engineering. We want fewer empty searches; use the figures in the draft.”
- Clarify audience and review depth. Read the evidence gaps and contradictions, keeping simulated perspectives separate from real research.
- Choose which fixes to draft. Persistent context and decision logging are separate choices; session-only answers are not saved to files.
FAQ
Source
Made by James Hemson. Original skill: pm, published on GitHub under the Apache-2.0 licence.
View on GitHub