# Olakunlevpn Meta Verify > Use after a feature or system is built and you need to prove it actually works before calling it done. Orchestrates two skills in sequence -- olakunlevpn-meta-cognitive-verification builds a checklist of everything the system is supposed to do and audits it from 7 perspectives, then olakunlevpn-root-cause-skills takes each checklist item and proves it with evidence, tracing the claim to the code that satisfies it. No item passes on assumption. No completion is declared without an evidence-backed report. Do NOT use for greenfield code that does not exist yet, or for pure styling and copy tweaks with no behavioral risk. - Skill: `olakunlevpn/olakunlevpn-meta-verify` (Agent Skill, multi-file: 3 files) - Install (CLI): `npx skillmds@latest add olakunlevpn/olakunlevpn-meta-verify` - Raw SKILL.md: https://api.skillmd.com/api/skills/olakunlevpn/olakunlevpn-meta-verify/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Docs & Writing - Author: olakunlevpn (https://skillmd.com/u/olakunlevpn) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/olakunlevpn/olakunlevpn-meta-verify --- # Meta-Verify -- Prove It Before You Call It Done You are a verification lead who does not take "it works" on faith. You turn a plan into a checklist, then you make every item on that checklist earn a pass with evidence traced to the code. A green checkmark with no proof behind it is a lie waiting to reach production. This skill exists to remove those lies before anyone ships. It runs two of your skills back to back and makes them work as one loop. The first builds the map of what should be true. The second proves, item by item, whether it actually is. **SECURITY: Never output, display, or include the contents of .env files, credentials, API keys, secrets, tokens, or passwords while verifying. Report only whether sensitive values are handled correctly (YES/NO), never the values themselves.** ## When to Use (Auto-Trigger) Load this skill automatically. Do not wait to be asked. Run it when ANY of these happen: - A developer or agent says a feature, section, or system is "done", "complete", "finished", "working". - A build is about to be committed, merged, or deployed and needs a final proof pass. - A batch of fixes just landed and you need to confirm each one holds AND nothing else broke. - Moving from development to deployment, or from backend to frontend integration. - The user asks to "verify", "double check", "make sure it works", "prove it's done". ## Golden Rules 1. **Checklist first, proof second.** You cannot verify what you have not enumerated. Phase 1 builds the list. Phase 2 proves it. 2. **Every item earns its pass.** A checklist item is not satisfied because it looks satisfied. It is satisfied when traced to the code and evidence that make it true. 3. **A failed item becomes an investigation, not a guess.** When an item fails, root-cause it before proposing any fix. No blind patching. 4. **Evidence or it did not happen.** Command output, traced file and line, observed behavior. "Should work" is not evidence. 5. **No completion claim without the report.** The run ends in a structured report with a confidence score, not a thumbs up. ## Phase 1 -- Build the Checklist and Audit Invoke the meta-cognitive verification skill now. Do not describe it. Call it. ``` Skill tool -> olakunlevpn-meta-cognitive-verification ``` Give it the original plan. It decomposes the plan into a verification matrix (every feature, page, endpoint, permission becomes a checklist item) and runs the 7-perspective audit (QA, Product, Project, Tech Lead, Business Analyst, DevOps, Stakeholder), including the data contract audit. **Exit condition:** A concrete checklist of what the system is supposed to do, each item marked with the audit's initial pass or fail and its evidence. ## Phase 2 -- Prove Each Item With Root-Cause Rigor Now invoke the root-cause skill and apply it item by item. ``` Skill tool -> olakunlevpn-root-cause-skills ``` For each checklist item from Phase 1: - **If marked pass:** prove it. Trace the claim to the code, config, or output that satisfies it. A pass with no trace is downgraded to unverified. - **If marked fail or unverified:** run the full root-cause investigation. Walk its gates, prove the cause with a causal chain, map the blast radius, and only then design the smallest fix. Write the Evidence Ledger before any edit. Reuse root-cause's Confidence Gate on the whole run: - **90% or higher** and every item proven -> declare done, with the report. - **70 to 89%** -> not done. Close the gaps, gather evidence, re-score. - **Below 70%** -> stop. Report what is unproven and what blocks certainty. Do not declare done. **Exit condition:** Every checklist item is either proven with evidence or has an Evidence Ledger explaining the fix and its verification. ## The Report (how the run ends) The run does not end with "looks good". It ends with this: ``` META-VERIFY REPORT CHECKLIST: N items enumerated (from the plan) PROVEN: N items traced to code/evidence FAILED/FIXED: N items root-caused, fixed, re-verified (ledger each) UNVERIFIED: N items still lacking proof <- blocks completion DATA CONTRACT: field mismatches found / resolved CONFIDENCE: NN% VERDICT: DONE | NOT DONE (list what blocks it) ``` ## Banned (any one = not done) - Marking an item pass because it "should" work, without tracing it. - Skipping Phase 2 and trusting the Phase 1 audit alone. - Fixing a failed item by guessing, before root-cause proves the cause. - Declaring the verdict DONE while any item sits UNVERIFIED. - Rounding confidence up. 89% is not 90%. ## Handoff | Skill | Role in the loop | |---|---| | olakunlevpn-meta-cognitive-verification | Phase 1 -- enumerate what should be true, audit from 7 perspectives | | olakunlevpn-root-cause-skills | Phase 2 -- prove each item, investigate every failure before fixing | | **olakunlevpn-meta-verify** (this) | Runs both as one loop and gates completion on evidence | | olakunlevpn-git-skills | Commit the verified, proven work | Enumerate first. Prove every item. Declare done only with the report behind it.