Review

Summary

Use this skill for static content quality reviews that produce clear findings and practical improvements.

When to use

• Reviewing documents for clarity, correctness, and completeness.
• Reviewing code for duplication, maintainability, and readability risks.
• Reviewing datasets for obvious consistency or redundancy issues.
• Providing a second opinion before publishing or merging content.

When not to use

• Runtime debugging or execution failures that require running systems or logs.
• Security audits or compliance checks that need specialized security workflows.
• Tasks requiring direct implementation rather than analysis.
• Reviews that depend on unavailable external business context.

Inputs

• Content to review (document, code, or data).
• Requested focus (for example: clarity, duplication, completeness, consistency).
• Optional context: style guide, acceptance criteria, constraints, or target audience.

Outputs

• Structured review report with:

1. Key findings (prioritized)
2. Duplications
3. Clarity/completeness gaps
4. Accuracy/consistency issues
5. Actionable improvements
6. Open questions (if context is missing)

Procedure

1. Clarify scope only if review focus is ambiguous.
2. Inspect content for duplication, ambiguity, inconsistency, and omission.
3. Apply provided standards or default community conventions.
4. Produce a concise, prioritized report with actionable fixes.
5. Separate confirmed findings from context-dependent questions.

Best practices and constraints

• Keep findings specific and evidence-based.
• Distinguish objective issues from style preferences.
• Prefer actionable recommendations over generic commentary.
• Do not claim runtime behavior; this skill is static analysis only.

Gotchas and edge cases

• Missing context can make valid patterns appear inconsistent; surface as open questions.
• Very small snippets can hide dependencies; avoid overconfident conclusions.
• Domain-specific correctness may require expert validation beyond this review.

Examples

• "Review this requirements doc for missing acceptance criteria and ambiguity."
• "Review this module for duplication and maintainability risks."
• "Review this dataset description for consistency and clarity."

Acceptance criteria

• Report is structured, prioritized, and actionable.
• Findings clearly separate confirmed issues from open questions.
• Recommendations are tied to observed content, not speculation.

Notes for agents

• Ask at most one clarifying question when scope is unclear.
• Cite exact sections or lines when available.
• Keep output concise and decision-oriented.

Limitations

• Does not execute code or validate runtime behavior.
• May miss highly domain-specific issues without additional context.
• Feedback quality depends on the completeness of provided inputs.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: nathanfaucett
• Source: nathanfaucett/agents
• License: MIT

Install and usage instructions live in the source repository linked above.