---
title: Verification Gate
category: product
entity_type: skill
price: $15
canonical: https://forgehouse.ai/skills/verification-gate/
lang: en
hreflang_alt: https://forgehouse.ai/tr/skiller/verification-gate/
last_updated: 2026-06-20
---

# Verification Gate

> Automatic completion verification gate.

An automatic completion gate that runs before any task is declared done. It applies a nine-point verification checklist: build, runtime, edge cases, regression, visual proof, claim labeling, trust gaps, and automatic-fail detection, and returns a hard PASS/FAIL verdict so unverified or overclaimed work never ships.

## Use cases
- Final check before declaring a task or fix complete
- Pre-deploy quality gate ahead of production
- Automated gate before a pull request merge
- Phase-transition checkpoint in multi-phase projects
- Catching overclaim before a status report goes out
- Verifying a fix targets its specific symptom, not just HTTP 200

## Benefits
- Tasks can't be closed until live, symptom-level proof exists
- Overclaiming is caught by enforced VERIFIED/INFERRED/ASSUMED labeling
- Regressions surface before users do, with neighbor-feature checks
- Every completion carries a clear PASS/FAIL verdict and remediation action

## What’s included
- Nine-point checklist: build, runtime, edge cases, limits, regression, and more
- Specific-symptom verification patterns beyond a bare HTTP 200 check
- Before/after comparison and regression-guard command patterns
- Claim-discipline labeling for every assertion in the output
- Structured PASS/FAIL output format with per-item results
- Exception handling for config-only, migration, and documentation changes

## Who it’s for
Builders and teams who want an enforced final gate that blocks unverified, overclaimed, or regression-risk work from being called done.

## How it runs
"Done" is a verdict, not a feeling. Nine gates stand between a finished task and the right to claim it, and a single auto-fail condition blocks the claim outright:
1. Build gate first: type check, production build and lint must all come back with zero errors; cheap checks run before expensive ones, and the first fail already decides the verdict.
2. Runtime gate on the live system: the production URL is hit for real, and not just for a 200 response; the specific symptom the fix claims to solve is what actually gets tested.
3. Edge cases: at least two alternate scenarios run every time, like empty data, a logged-out user or a different language.
4. Regression and visual pass: neighbouring features get re-checked so the fix didn't break what already worked, and UI work gets screenshots at three breakpoints.
5. Claim audit: every assertion is tagged as verified, inferred, assumed or not verified; an unproven claim stated in confident language counts as an automatic fail.
6. Final verdict: a line-per-item PASS/FAIL report; any single auto-fail condition (no rollback path, untested production change, output that looks good but cannot be measured) blocks the done claim until fixed.

## FAQ
### I'm a solo developer shipping small fixes, isn't a nine-point gate overkill?
The gate scales to the change. Config-only, migration, and documentation changes go through explicit exception handling instead of the full checklist, while anything touching runtime behavior still gets symptom-level proof. Solo work is where unverified 'done' claims slip through most easily.

### How does it actually catch overclaiming instead of trusting my word?
Two mechanisms. Every assertion must carry a VERIFIED, INFERRED, or ASSUMED label, and verification targets the specific symptom the fix claims to solve rather than a bare HTTP 200. If live proof is missing or a label is inflated, the verdict is FAIL with a remediation action.

### Once the verdict comes back FAIL, will the gate also write the remediation for me?
No. The gate blocks the completion claim and returns a per-item PASS/FAIL with what's missing, but writing the fix and re-running the gate is your job. It's a referee, not a repair crew.

## Price
$15, one-time, no subscription. VAT included.

Related guide: [How to run a marketing agency with AI automation](https://forgehouse.ai/guides/ai-marketing-agency-automation/)