Catch the accessibility failures that ship in almost every AI-built UI. Use after building any interactive component.
中文简介正在准备中。当前显示原文说明。
Use before claiming work is complete, fixed, or passing — before committing, opening a PR, or handing off. Requires running the verification command in THIS turn and reading its output before any success claim.
中文简介正在准备中。当前显示原文说明。
查看安装方法安装前可以查看交给 AI 代理的指令内容。
Claiming work done without fresh verification is dishonesty, not efficiency. adversarial-verify is the what; this skill is the when — the gate you pass through right before any completion claim.
NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE
If you have not run the verification command in this message, you cannot claim it passes. Not "should", not "probably", not "based on the diff".
Before writing "done" / "fixed" / "green" / "ready to merge" — even in your own head:
Skip any step = you are lying to the user, not verifying.
| Claim | Requires | Not sufficient |
|---|---|---|
| Tests pass | Fresh test run, exit 0, 0 failures | "should pass", previous run, "logic looks right" |
| Linter clean | Linter output, 0 errors | Partial check, extrapolating from unrelated files |
| Build succeeds | Build command, exit 0 | Linter passing, editor squiggles gone |
| Bug fixed | Reproduce original symptom, watch it not happen | Code changed, "assumed" fixed |
| Regression test works | Red → green cycle verified (revert fix, watch test fail, restore, watch pass) | Test passes once |
| Agent/subagent completed | Read the VCS diff, verify claimed changes exist | Agent's own "success" report |
| Spec satisfied | Line-by-line checklist against the plan | "Tests pass, phase complete" |
| Excuse | Reality |
|---|---|
| "Should work now" | RUN it. |
| "I'm confident" | Confidence ≠ evidence. |
| "Linter passed" | Linter ≠ compiler ≠ tests. |
| "The agent said success" | Read the diff yourself. |
| "Partial check is enough" | Partial proves nothing about the whole. |
| "Different words, so rule doesn't apply" | Spirit over letter. |
Tests
34/34 pass. Then say "all tests pass".Regression tests (real red-green)
Build
Agent delegation
Always, before:
adversarial-verify — the 11 shortcuts agents take to fake "done"; run through the list, then run through this gate.clean-commits — clean commits require verified content.verifier subagent — dispatch it; then verify its report against the diff (per the "Agent delegation" pattern above).Run the command. Read the output. THEN claim the result. Non-negotiable.
还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
Catch the accessibility failures that ship in almost every AI-built UI. Use after building any interactive component.
中文简介正在准备中。当前显示原文说明。
Before compaction Loopkit extracts decisions into claude-decisions.json (machine-readable). Read it alongside claude-progress.txt at session start — prose is for humans, JSON is for the loop.
中文简介正在准备中。当前显示原文说明。
Review a diff against the goal spec assuming the code is BROKEN. The reviewer that lives in the maker's head always agrees with itself — this pulls review into a hostile, separate pass. Invoke after every code change before marking work done.
中文简介正在准备中。当前显示原文说明。
Verify that an endpoint checks ownership, not just authentication. Use on any handler that reads or mutates user data.
中文简介正在准备中。当前显示原文说明。
Find the exact commit that introduced a bug. Use when something worked before and broke, and you don't know which change did it.
中文简介正在准备中。当前显示原文说明。
Before picking new work, smoke-test the last "completed" feature. If it's broken, revert and re-open it before touching anything else. Kills the "looks shipped, isn't shipped" bug across sessions.
中文简介正在准备中。当前显示原文说明。