菜单
Prove that work actually behaves correctly by spawning the verifier agent to run the code and capture evidence. Use before claiming a task complete, before commit, when in doubt about whether tests prove the right thing. Triggers on /verify, "prove it works", "verify this", "did the change actually work".
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/sethdford/claude-agent-os --skill verify命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Run the bounded harness caretaker — deterministic state gather (predictions, verifier gain, agent health, sensors, drift), then triage ONLY what's flagged under a hard authority boundary (files work, never ships changes). Triggers on /caretaker, "run the caretaker", "harness health check".
Mine past Claude Code session transcripts (JSONL) to extract user corrections, successful patterns, and recurring failure modes. Proposes diffs against lessons.md and rules/*.md for human review. Triggers on /mine-transcripts, "mine my sessions", "what did I learn this week", "session retro". Use weekly or after a fleet completes.
Apply the diff/patch outputs from a /mine-transcripts run to lessons.md and rules/*.md after human review. Use after running /mine-transcripts and reviewing the proposed diffs. Triggers on /apply-mining-patches, "apply the mining patches", "apply mining run", "approve and apply the trajectory diffs".
基于 SOC 职业分类
正在显示 SKILL.md
| name | verify |
| description | Prove that work actually behaves correctly by spawning the verifier agent to run the code and capture evidence. Use before claiming a task complete, before commit, when in doubt about whether tests prove the right thing. Triggers on /verify, "prove it works", "verify this", "did the change actually work". |
| when_to_use | Whenever you are about to assert a task is done, a fix is in, or a behavior holds. The verifier runs the code; you do not assert from reading. |
| allowed-tools | Agent, Read, Bash |
| arguments | ["target"] |
/verify [target] spawns the verifier agent to actually execute the code and capture evidence.
Identify the target. If target is given, that's the change/area to verify. Otherwise, infer from recent context: the file you just edited, the task you just claimed done, the bug you just fixed.
Compile the contract. List 3–8 testable behaviors the change is supposed to satisfy. If you can't list them, the work isn't ready to verify — clarify the contract first.
Spawn the verifier. Use the Agent tool with subagent_type: verifier. Pass:
pnpm test, pytest path/to/file.py, curl ...)Read the verifier's output. It will end with one of:
RESULT_verifier=PASS — proceed with whatever you were about to doRESULT_verifier=FAIL — STOP. Surface the failed behaviors. Do not claim done.RESULT_verifier=INCONCLUSIVE — STOP. Surface the blocker. Verification couldn't complete.Never override. If the verifier says FAIL or INCONCLUSIVE, you do not say "I think it's fine anyway." You either fix the issue or surface it to the user.
/eval: run /verify on any newly-built skill before promoting itcommit-commands:commit: /verify the change first/team: every agent's task closure runs /verify before TaskUpdate marks it done