SuperAPI AI Directory

thermo-fix

Testing & quality

View repository

Updated 10-15 21:09

About this skill

agentic-workflows-blueprint.workflow.thermo-fix

Goal

Run a thermo-nuclear code quality review on the diff, apply every identified fix inline, verify the build, and iterate until all blockers are resolved (up to 3 cycles).

Scope

  • Applies to: Files modified in the current session or branch.
  • Does not cover: Implementing new feature scope from scratch or running deployment pipelines.

Triggers

  • "/thermo-fix"
  • "Run review and fix cycle"
  • "Fix code quality findings"

Inputs

  • diffScope: Files to review (default: modified files in branch).
  • maxIterations: Maximum review+fix cycles (default: 3).
  • buildCommand: Project verification/build command.

Invariants

  1. Never approve on behavior alone: Structural regressions are blockers.
  2. Inline fix cycle: Apply fixes in the same conversation without branching.
  3. Build verification: Run build/syntax check after every fix cycle.
  4. No unsolicited changes: Only fix identified quality findings.

Procedure

  1. Identify Diff: List all modified files.
  2. Review: Run thermo-nuclear-code-quality-review to produce finding table (Blocker, High, Medium, Low).
  3. Apply Fixes: Fix findings top-down (Blocker -> High -> Medium).
  4. Verify Build: Run project build/check command. Fix build errors if any occur.
  5. Re-review: Re-inspect all changed files. Stop if zero Blocker or High findings remain. Otherwise repeat up to maxIterations.
  6. Commit: Create a local git commit with finding summary.

Outputs

  • Resolved code findings.
  • Build verification pass.
  • Updated docs/tasks/task.<reference>.md and walkthrough.md.
  • Local git commit.

Review gate

  • Zero Blocker findings.
  • Zero High findings.
  • Project build/syntax checks pass.

References

  • ../../SKILL.md
  • ../thermo-nuclear-code-quality-review/SKILL.md
  • ../radioactive/SKILL.md