Evaluate codebase architecture for modularity, coupling, and misplaced boundaries, then write an audit report.
npx skills add https://github.com/kieranpotts/skills --skill audit-kieranpotts
Please help me install this Agent Skill. Skill: audit Source: https://github.com/kieranpotts/skills/tree/main/skills/audit Command: npx skills add https://github.com/kieranpotts/skills --skill audit-kieranpotts
High-intent search queries and answers about installing and using this skill.
Invoke the audit with a phrase like "audit this codebase" or "do an architectural audit". You can name a target path or repository URL and pin a specific commit; otherwise the current repository at its checked-out revision is audited.
The audit checks module depth, boundary placement and leakage, tangled dependencies, single-caller abstractions, inverted dependencies, and misnamed abstractions. Every finding must cite specific files and lines, and findings are prioritized by impact divided by effort.
No. The audit is discovery-only: the audited codebase must remain byte-for-byte unchanged, with no commits, branches, issues, or pull requests raised. The only output is a single report written to the resolved audit reports store.
No. Security, privacy, and runtime performance findings are explicitly out of scope and must not appear in the report. Those concerns are handled by separate exercises such as a security probe with its own methods.
Deliberate blindness to design docs and decision records keeps the review unbiased, so the agent forms judgment from the code and data structures alone. The trade-off is some noise from retreading settled trade-offs.
Avoid running it on every commit, especially in continuous integration, since it is open-ended analysis best suited to a premium reasoning model. Schedule it periodically, after large changesets, or as a release preflight step.
udecodeMaintains Plate's latest and beta npm release lanes across main and next branches.
JBonfimDetect code smells and map them to Fowler-style refactoring techniques.
Agile-VPerform STRIDE threat modeling and privacy impact assessments to generate security requirement artifacts.
AndurilCodeExtract business logic and domain models into an llms.txt-style Markdown reference.
