Lean instruction-only body that assumes Claude's competence — a single orienting sentence then straight to a curl command and checklists, with no padding or explanation of basic concepts; every line earns its place.

Provides copy-paste-ready executable commands ('curl -s "https://transparencyreport.google.com/..."', 'curl -sL https://the-domain/') plus concrete bad/good domain examples ('youtube-x.com', 'tool.yourname.com') and a fully worked real example.

Prevention, Diagnosis, and Recovery are numbered sequences with an explicit validation checkpoint ('re-run the Quick check until it returns STATUS 1, then confirm in a browser') and a feedback loop (fix surface, deploy, then request review), satisfying the destructive/batch-operation validation requirement.

Single-file skill (~45 lines) with no bundle directories and well-organized sections (Quick check, Prevention checklist, Diagnosis workflow, Recovery, Worked example); per the simple-skills note, this clears the bar without external references.

Names concrete actions ('Prevent and fix Google Safe Browsing flags') plus several specific triggering scenarios (launching a public web app, buying/picking a domain, building a login/signup page), matching the 'lists multiple specific concrete actions' anchor.

Clearly answers both what ('Prevent and fix Google Safe Browsing Dangerous site flags') and when ('Use when launching a public web app... or when any site shows a red warning'), with an explicit trigger clause; not level 2 because the 'when' is explicit rather than implied.

Explicit 'Triggers on' clause surfaces natural user phrasing — 'dangerous site', 'deceptive site', 'site blocked', 'safe browsing', 'phishing flag', 'red warning screen' — giving good coverage of terms a user would actually say.

Occupies a clear niche (one upstream blocklist feeding all major browsers) with distinctive browser-warning triggers unlikely to fire for unrelated skills.