# Verification Protocol

> Producer vs evaluator separation. S5 Execution implements; S6 Forensic Review verifies. Do not rely on self-critique alone.

- Skill: `tools-only/verification-protocol` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add tools-only/verification-protocol`
- Raw SKILL.md: https://api.skillmd.com/api/skills/tools-only/verification-protocol/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: tools-only (https://skillmd.com/u/tools-only)
- Updated: 2026-09-29
- Page: https://skillmd.com/skills/tools-only/verification-protocol

---

# Verification Protocol

Producer vs evaluator separation. S5 Execution implements; S6 Forensic Review verifies. Do not rely on self-critique alone.

---

## Verdicts

| Stage | Verdicts |
|-------|----------|
| S6 Forensic Review | COMPLETE / NEEDS_WORK |
| S7 Final Verification | CERTIFIED / NOT_CERTIFIED |

---

## verification-gate (Pre-S5)

Before S5 Execution self-verification, run 4 checkpoints before any Bash/Write/Edit/NotebookEdit:

1. **Hypothesis Stated** — Explicitly state what system/component the issue affects
2. **Hypothesis Verified** — Gather evidence (Read, Grep, MCP tools)
3. **Hypothesis-Action Alignment** — Action targets same system as hypothesis
4. **Pattern-Matching Detection** — Verified against project reality, not training data

---

## verify Skill (Completion Checklist)

Before claiming task complete:

- Task type & strategy
- "WORKS" check (executable or static)
- "FIXED" check (for bug fixes)
- Quality gates (pre-commit, lint, typecheck)
- Honesty check

**Golden rule**: If you cannot demonstrate it working with evidence, it is NOT done.

---

## Sources

- [verify SKILL.md](../../.claude/skills/verify/SKILL.md)
- [verification-gate SKILL.md](../../plugins/verification-gate/skills/verification-gate/SKILL.md)