Install
$ agentstack add skill-yihannangua-prd-review-skill-prd-review ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
PRD Review
This skill performs a deep PRD and requirements quality audit. It is for understanding and challenging the business logic in a requirement document, not for checking whether a document follows a fixed template.
Core Intent
Review whether the requirement is correct, clear, consistent, complete, implementable, and testable.
Before criticizing, build a business understanding:
- Business goal and system boundary
- Actors, roles, and permissions
- Business objects and data relationships
- Module responsibilities and module boundaries
- Core workflows and state transitions
- Cross-module dependencies and side effects
- Normal, exception, boundary, retry, rollback, and audit rules
Distinguish explicitly between:
- Stated: directly present in the source document
- Inferred: reasonable inference from the document
- Missing: required for implementation or testing but absent
- Question: must be confirmed by product/business owner
When To Use
Use this skill when the user asks to:
- Deeply review a PRD or requirement document
- Find business logic errors, omissions, contradictions, or ambiguity
- Evaluate module relationships, internal rules, and cross-module flows
- Check whether developers and testers can implement and verify the requirement
- Produce detailed modification advice, not just a score
Do not use this as the primary skill when the user only wants grammar cleanup, document formatting, or a fixed delivery checklist review.
Required Workflow
- Ingest source material
- Read the user-provided document(s), preserving section names and locations.
- For DOCX/PDF inputs, extract text with the appropriate document tooling.
- Note unreadable diagrams, external links, missing attachments, or image-only flowcharts.
- Load review framework
- Read
references/review-framework.mdbefore the first substantive review. - Read
references/issue-taxonomy.mdbefore classifying findings. - Read
references/report-template.mdbefore writing the final report. - Read additional reference files only when needed:
references/module-review-template.mdfor per-module review.references/cross-module-checklist.mdfor module interaction and consistency checks.references/business-rule-checklist.mdfor rules, states, data, permission, and exception checks.references/requirement-quality-checklist.mdfor clarity, consistency, priority, terminology, data definition, and ambiguity checks.
- Build business model first
- Summarize the business goal, system boundary, actors, objects, modules, flows, states, data dependencies, and permissions.
- Mark each important point as Stated, Inferred, Missing, or Question.
- If the business model cannot be built, report blockers before issuing detailed findings.
- Perform module-level review
- For each module, identify responsibilities, inputs, outputs, dependencies, rules, states, permissions, normal flow, exception flow, and missing decisions.
- Find internal contradictions, underspecified rules, incomplete states, ambiguous wording, and test gaps.
- Check whether business rules are explicitly defined and whether requirement descriptions are unambiguous enough for design, development, and QA.
- Perform cross-module review
- Check whether upstream data exists before downstream use.
- Check consistency of field definitions, conflict handling, permissions, retries, status names, and side effects.
- Check consistency of terminology, feature descriptions, data definitions, priority labels, and priority criteria.
- Check whether data changes in one module affect other modules and whether those effects are specified.
- Assess implementation and testing risk
- Identify rules that cannot be implemented deterministically.
- Identify acceptance criteria or test cases that cannot be derived from the current document.
- Identify external system, data migration, concurrency, audit, rollback, and security risks.
- Write findings with fixes
- Every important finding must include why it matters, impact, specific recommendation, and a suggested replacement/addition when useful.
- Do not merely say "missing acceptance criteria"; name the concrete scenario and propose testable wording.
- Put open business decisions into a separate "Questions For Product" section.
Output Expectations
The final report should be actionable for product, engineering, and QA.
Include:
- Overall conclusion: whether the requirement is ready for design/development/testing.
- Business understanding: concise model of the domain as read from the document.
- Key risks: the highest impact logic, data, permission, or testability risks.
- Module-level findings.
- Cross-module findings.
- Detailed findings table.
- Suggested document changes, including suggested wording for high-value fixes.
- Questions for product/business owner.
- Suggested test scenarios.
Scoring is optional. If a score is useful, keep it secondary to the findings and rationale.
Review Discipline
- Do not invent facts. Label inferences.
- Do not stop at template gaps. Find business and rule gaps.
- Do not treat screenshots or diagrams as sufficient when the underlying rules are not specified.
- Do not require a single universal template; adapt the review to the document's domain.
- Prefer precise, cited findings over broad criticism.
- Use source section names, line numbers, headings, or quoted short phrases to locate issues.
- Keep the tone firm, specific, and constructive.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: yihannangua
- Source: yihannangua/prd-review-skill
- License: MIT
- Homepage: https://github.com/yihannangua/prd-review-skill
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.