Survey scoped technical debt using the repository's existing analyzers and render a prioritized, evidence-grounded report without implementing fixes. Use for technical-debt reviews of changed files, a component, a work item, or the whole repository.
Recursos
1Instal·la
npx skillscat add tvproductions/gz-skills/gzs-tech-debt-review Instal·la des del registre de SkillsCat.
GovZero Technical-Debt Review
Survey debt; do not patch it. Wield existing project probes, preserve their raw
evidence, and recommend a concrete fix shape and route for every retained
finding.
Discovery and fallback
Prefer, in order:
- The repository's own probe wrappers, which already encode its thresholds.
- The underlying analyzers the repository configures, invoked directly.
- Direct reading of the resolved file set, plus repository history for churn
and staleness signals.
The third rung is the floor and needs only the source and its history. It
yields fewer categories, so name the ones it cannot reach instead of dropping
them silently. A selected category with no available probe is itself a finding,
not an omission.
Workflow
Resolve a bounded scope from the user's request: changed files, explicit
paths, a work item, or the whole repository. Save or report the exact file set
so the audit can be repeated.Discover available canonical probes. Use only evidenced categories, such as
complexity and size, lint and types, test gaps, dead code, dependency drift,
documentation drift, stale TODOs, packaging boundaries, portability, or
generated-surface coherence. Prefer repository wrappers over rebuilding
their command sequences.Run every selected probe and retain its command, exit status, and raw output
location. A missing or broken probe is itself a finding; never silently drop
a category after selecting it.Convert probe hits into findings with class, severity, location, verbatim
evidence, recommended fix shape, and route. Remove verified false positives
with a recorded reason.Grade severity by blast radius and reversibility, not by estimated diff size:
- Critical: violates a published contract, security boundary, or release gate.
- High: breaks binding project policy or blocks the next change on the surface.
- Medium: compounds silently but does not block near-term delivery.
- Low: local, cosmetic, or safely autotool-fixable drift.
Recommend an imperative fix shape, not “review,” “clean up,” or speculative
new capability. Route through the repository's existing maintenance or issue
process, but do not create external work items without authorization.Render a severity-sorted summary and cited detail for Critical and High
findings. End with the smallest useful next action.
Evidence
Report the scope, probes and failures, severity counts, highest-risk findings,
report path when written, and recommended routes. The review is incomplete when
a selected probe has no recorded outcome.