01Why it exists
A finished page, a passing build, or one HTTP 200 each proves only one layer. This Skill separates the evidence often conflated before release, so a team can see when to stop and hand the decision to a person.
An offline release-review Skill; it does not deploy sites, change DNS, or operate accounts or platforms.
An open-source Skill that uses six evidence layers and stop gates to assess whether an AI-assisted website project may enter human release review.

A finished page, a passing build, or one HTTP 200 each proves only one layer. This Skill separates the evidence often conflated before release, so a team can see when to stop and hand the decision to a person.
The English public Skill is on main of 970lab/ai-content-ops at bb770b4cd4155165aa0c85286b67f46d5acf49c9, under skills/ai-website-release-gate/. It provides offline, evidence-based candidate, blocker and no_candidate outcomes; it does not deploy a site, change DNS, submit indexing, or operate accounts or platforms.
Research · Planning · Content · Testing · Automation
General method, public repository code and fictional examples only; no internal remediation matrices, private paths, accounts, client, family or child information, credentials, internal topology, or unverified runtime state.
After visual-asset provenance, language, de-identification and usage scope are approved, separately authorise site change, then verify Preview, Production and the formal domain before treating this site page as public-release fact.