API Review Skill for Claude

API Review Workflow

When NOT To Use

• Internal refactors behind a stable surface (use

pensive:code-refinement)

• Coupling and layering questions (use pensive:architecture-review)

Table of Contents

• Usage
• Required Progress Tracking
• Workflow

Usage

Use this skill to review public API changes, design new surfaces, audit consistency, and validate documentation completeness. Run it before any API release to confirm alignment with project guidelines.

Required Progress Tracking

• api-review:surface-inventory
• api-review:exemplar-research
• api-review:consistency-audit
• api-review:docs-governance
• api-review:evidence-log
• api-review:findings-verified

Workflow

Step 1: Surface Inventory

Catalog all public APIs by language. Record stability levels, feature flags, and versioning metadata. Use tools like rg to find public symbols (e.g., pub in Rust or non-underscored def in Python). Confirm the working tree state with git status before starting.

Step 2: Exemplar Research

Identify at least two high-quality API references for the relevant language, such as pandas, requests, or tokio. Document their patterns for namespacing, pagination, error handling, and structure to serve as a baseline for the audit.

Step 3: Consistency Audit

Compare the project's API against the identified exemplar patterns. Analyze naming conventions, parameter ordering, return types, and error semantics. Identify duplication, leaky abstractions, missing feature gates, and documentation gaps.

Step 4: Documentation Governance

Validate that documentation includes entry points, quickstarts, and a complete API reference. Verify that changelogs and migration notes are maintained. Check for SemVer compliance, stability promises, and clear deprecation timelines. Confirm that documentation is generated automatically using tools like rustdoc, Sphinx, or OpenAPI.

Step 5: Evidence Log

Record all executed commands and findings. Summarize the final recommendation as Approve, Approve with actions, or Block. Include specific action items with assigned owners and due dates.

API Quality Checklist

Naming

Confirm consistent conventions and descriptive names that follow language-specific idioms.

Parameters

Verify consistent ordering and ensure optional parameters have explicit defaults. Check that type annotations are complete.

Return Values

Analyze return patterns for consistency. Confirm that error cases are documented and that pagination follows a uniform structure.

Documentation

Verify that all public APIs include usage examples and that the changelog reflects current changes.

Output Format

The final report must include a summary of the API surface, a numerical inventory

of endpoints and public types, and an alignment analysis against researched

exemplars. Document consistency issues and documentation gaps with precise file

and line references. Conclude with a clear decision and a timed action plan.

Each issue must follow this structure:

[A1] Title
- Location: file.py:42
- Anchor: `verbatim source text at line 42`
- Issue: what is wrong | Fix: remediation | Evidence: [E1]

The Anchor is the exact source text at Location; it is what

citation_verifier.py re-reads to prove the finding is real.

Technical Integration

Use imbue:proof-of-work for reproducible command capture and imbue:structured-output for formatting findings. Reference imbue:diff-analysis/modules/risk-assessment-framework when assessing breaking changes.

Module Reference

• See modules/surface-inventory.md for API cataloging patterns
• See modules/exemplar-research.md for researching API standards
• See modules/consistency-audit.md for cross-API consistency checks

Verify Findings Are Grounded (api-review:findings-verified)

Every finding must cite a real location and a verbatim anchor. Write

findings to .review/findings.json and confirm each citation resolves:

python plugins/imbue/scripts/citation_verifier.py \
  --findings .review/findings.json --repo-root .

Drop or label UNVERIFIED any finding the verifier fails (exit 1); only

verified findings enter the report. See Skill(imbue:review-core) Step 5

and Skill(imbue:structured-output) for the schema.

Exit Criteria

• Surface inventoried, exemplars researched, consistency audited,

documentation governance checked, and evidence logged.

• Every reported finding carries a Location + verbatim Anchor

confirmed by citation_verifier.py (exit 0), or unverified findings

were dropped or labeled UNVERIFIED.

Troubleshooting

If the audit command is missing, verify that dependencies are installed and

accessible in the system PATH. Check file permissions if access errors

occur. Use the --verbose flag to inspect execution logs if the tool

behaves unexpectedly.