test-auto-fix

Autonomously diagnose and fix failing tests in any project, with hard guardrails so the loop never devolves into "make the red go away".

This SKILL.md is the orchestration index. Load the matching rule file when you need detail — do not preload them.

Phase	Goal	Required rule
0	Resolve the surface (bootstrap or validate)	`rules/bootstrap.md` + `rules/surface-validation.md`
1	Detect which test surface(s) are failing; build fix plan	this file + `templates/plan-artifact.md`
2	Per failure: classify as test-bug / prod-bug / unsure	`rules/verdicts.md` + `rules/self-improvement-loop.md` (read lessons)
3	Per failure: draft the smallest possible fix	this file
3.5	Confidence gate — before any edit	`rules/confidence-gate.md`
4	Apply + verify single failing test	this file + `rules/anti-patterns.md`
5	test-provenance-guard (optional companion)	invoke `Skill("test-provenance-guard")`
6	Outer loop: re-run full surface; regression-detect	`rules/regression-detection.md` + `rules/self-improvement-loop.md` (write lessons)
7	Report (structured exit summary)	`templates/exit-summary.md` + `rules/self-improvement-loop.md` (write on outcome)