From loopkit
Reviews diffs against the goal spec assuming code is broken, checking 11 shortcuts agents use to fake done. Outputs JSON verdict.
How this skill is triggered — by the user, by Claude, or both
Slash command
/loopkit:adversarial-verifyWhen to use
a code change is "done", before flipping a task to complete, before commit
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Default stance: **the code is broken until proven otherwise.** Your job is to find where.
Default stance: the code is broken until proven otherwise. Your job is to find where. Do not be polite. Do not propose fixes. Do not run the code. Just hunt.
{"passes": false, "failures": [{"line": 42, "shortcut": "swallowed errors", "why": "..."}]}
If it genuinely passes, say so in one line. Most of the time, it doesn't.
npx claudepluginhub archive228/loopkitMulti-pass structural code review with adversarial analysis, scope-aware dispatch, and forced-finding protocol. Supports branch diff, PR number, commit hash, and pending changes.
Static code review of the active diff: traces changed paths and reports P1/P2/P3 correctness, security, or spec bugs with file:line evidence. Use for code review or bug checks.
Reviews pull requests, branch diffs, and local working-tree diffs for correctness, security, concurrency, performance, and code quality issues. Returns structured JSON findings.