tech-debt

Scans and prioritizes tech debt and code-health cleanups. Use when asking what to clean up, assessing code health, planning refactors, or prioritizing debt. Do not use for a single design decision (ar

SKILL.md

GitHub

Tech Debt

Surface debt, rank it, and propose small reversible cleanups — do not boil the ocean.

Related

  • Deepening opportunities with domain language → improve-codebase-architecture
  • Single design choice → architecture

Workflow

  1. Scope — directory, service, or theme (testability, duplication, coupling).
  2. Collect signals — hotspots, shallow modules, duplicated logic, missing tests, TODO/FIXME age, flaky tests, oversized files. See references/debt-signals.md.
  3. Rank — impact × frequency × risk of change. Prefer high-leverage, low-blast-radius items.
  4. Propose slices — each item: problem, proposed move, verification, estimated risk. One PR-sized slice per item when possible.
  5. Separate from features — do not mix opportunistic cleanup into unrelated ticket PRs unless required for the change.

Constraints

  • Refactors preserve behavior by default unless the goal explicitly changes behavior.
  • Prefer deleting pass-through abstractions over polishing them.
  • No speculative generality “for later”.

Verification

  • Scope stated
  • Ranked list with rationale
  • Each proposed slice has a verification plan
  • Non-goals / deferred items listed