Verify+

verify+ — Evidence-Based Verification

Wraps superpowers:verification-before-completion. Requires superpowers to be installed.

This skill activates verify+ phase in the enforcement layer. During this phase:

• Full test suite runs are ALLOWED (this is the verification phase)
• Zero-defect enforcement is at strict tolerance
• Evidence must be shown before claiming done

Procedure

Phase A: Preparation

1. Announce phase entry:

``` Now using verify+ skill. Enforcement layer adjusted:

• Test scope: full suite runs allowed (this is verification phase)
• Zero-defect: strict — every failure must be fixed
• Evidence: show command output before claiming done

```

1. Load the plan from docs/plans/.

1. List all acceptance criteria from the plan.

Phase B: Full Verification (delegate to superpowers:verification-before-completion)

1. Invoke superpowers:verification-before-completion with the plan context.

1. Run the FULL test suite:

`` npx vitest run ``

Show the output. Every test must pass. Zero failures. Zero errors.

1. For each acceptance criterion in the plan:

• Show evidence that it is met (command output, file contents, test results)
• Use positive framing: "Criterion X is met: [evidence]"
• Do NOT say "tests pass" — show the actual test output

1. Check for spec drift:

• Compare the plan's task list against actual implementation
• Were any tasks skipped? Added? Changed?
• Document any deviations with reasons

Phase C: Evidence Report

1. Produce a verification report:

``` ## Verification Report

### Test Suite [Full test output — ALL must pass]

### Acceptance Criteria

• [ ] Criterion 1: [evidence]
• [ ] Criterion 2: [evidence]
• [ ] ...

### Spec Drift

• [No deviations / List deviations with reasons]

### Enforcement Compliance

• [ ] Active enforcement rules followed (see session-start output; real dependencies in stack/E2E tests by default)
• [ ] If the design defines a signal stack: every applicable layer's signal was run, with each layer's result reported (pass / diff / drift / trace)
• [ ] Evidence shown for all claims
• [ ] All source changes have test coverage

```

1. If ALL checks pass, verification is complete.

If ANY check fails, return to the implementation phase (tdd+ or sdd+) to fix.

Skill Chain

After verify+ passes:

• Invoke /review+ to run the compliance review agent

Completion

Report one of these states when the skill finishes:

• DONE — Full test suite passes, all acceptance criteria met with evidence, no spec drift.
• DONEWITHCONCERNS — All tests pass but minor spec deviations exist (documented).
• BLOCKED — Tests failing, cannot proceed without fixing regressions.
• NEEDS_CONTEXT — Need user input to resolve an acceptance criterion or spec drift question.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: franklywatson
• Source: franklywatson/claude-rig
• License: MIT

Install and usage instructions live in the source repository linked above.