# Goal Verifier

> Verify refactor-safely goals: full test lock, no public API breakage, behavior preserved. Use before update_goal(completed: true) on refactor goals.

- Skill: `cobusgreyling/goal-verifier-5` (Agent Skill)
- Install (CLI): `npx skillmds@latest add cobusgreyling/goal-verifier-5`
- Raw SKILL.md: https://api.skillmd.com/api/skills/cobusgreyling/goal-verifier-5/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Integrations & APIs
- Author: cobusgreyling (https://skillmd.com/u/cobusgreyling)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/cobusgreyling/goal-verifier-5

---


# Goal Verifier — Refactor Safely

Checker only. Do not implement fixes.

## Steps

1. Read `GOAL.md` — refactor scope, done conditions.
2. Run **full** test suite (not subset) from `AGENTS.md`.
3. If public API listed in scope, diff exports/signatures — REJECT on unintended breakage.
4. REJECT if test files were weakened (removed assertions, broadened matchers) without GOAL.md approval.
5. Return `VERDICT: PASS` or `VERDICT: REJECT`.

## Output

```
VERDICT: PASS | REJECT

Evidence:
- tests: [command, exit code, count]
- api check: [summary]

Gaps (if REJECT):
- [failing tests or API breaks]
```

## Rules

- Behavior-preserving means tests pass without changing expectations.
- REJECT drive-by feature additions outside refactor scope.