跳到正文
cccskills
免费在 GitHub 公开

power-verify

Use when about to claim work is complete, fixed, or passing, and when a finished feature needs adversarial verification before integration

中文简介正在准备中。当前显示原文说明。

查看安装方法

包含的文件(2)

  • SKILL.md3.4 KB
  • agents/openai.yaml189 B

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Evidence Before Claims

Read collaboration when dispatching the verifier — its status is at most ten lines and its report is never pasted into your context — and artifacts before writing verdict.md or state.md.

Route

  • no argument, mid-work: the gate below, then continue
  • <feature-slug>: full adversarial verification, ending in a verdict
  • --strict: two verifier lenses in parallel; the verdict is the worse of the two

The gate

Before any claim that something works, fixed, passes, or is done:

  1. Identify the command that would prove it. .planning/power/context/project.md records this repository's real test, lint and build commands; use those rather than the ones that are conventional for the ecosystem.
  2. Run it, fresh and complete. Not a cached result, not a subset.
  3. Read the whole output: the exit code, the failure count, the warnings.
  4. Verify the output actually says what you are about to claim.
  5. Only then, claim it.

Skipping a step is not verifying faster; it is asserting.

ClaimRequiresNot sufficient
"Tests pass"the full suite run now, output readthe last run, before your change
"It builds"the real build commandthe linter passed
"The bug is fixed"the regression test failing before and passing afterthe symptom is gone
"Nothing else broke"the full suitethe touched file's tests
"The agent did it"the VCS diffthe agent said DONE

No satisfaction before evidence. Not "Great!", not "Perfect!", not "That should do it". Those phrases commit you to a conclusion you have not yet checked, and having committed, you will read the output looking for agreement.

Adversarial verification

When verifying a whole feature, dispatch the verifier subagent. Its instruction is not "check whether this is complete" — it is "try to refute that this is complete".

Give it the spec, the plan, the ledger, and the branch range. It falsifies each stated truth — objective, acceptance criteria, definition of done, prohibitions — with real commands, and reports what survived.

--strict runs two lenses in parallel:

  • Functional: does it do what the spec says, under the spec's exact values?
  • Compliance: does it respect the prohibitions, conventions and non-functional constraints?

The final verdict is the worse of the two. A feature that works but violates a stated prohibition is not approved.

Verdicts

VerdictMeaningNext
APPROVEDEvery stated truth survived refutation.pwdev-power:power-finish
CAVEATSApproved, with findings recorded that nothing blocks on.pwdev-power:power-finish, caveats surfaced
REJECTEDAt least one stated truth was falsified.fix plan, then re-verify

Write the verdict to .planning/power/features/<slug>/verdict.md with the evidence — the commands run and their output — and record the gate in state.md.

Correction cap

A REJECTED verdict produces fix-<NN>.md, executed through pwdev-power:power-execute. At most two correction cycles. A third rejection stops and goes to the human; it never becomes an approval by attrition.

评价

还没有评价。欢迎分享你的使用感受。

评论

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。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Accessibility auditing and remediation skill for Vue 3 + Reka UI. Use when reviewing components, verifying WCAG 2.1 AA, fixing ARIA issues, or validating keyboard navigation in Reka UI components.

中文简介正在准备中。当前显示原文说明。

pwdev-solucoes/pwdev-claude-marketplace32026年10月15日 更新

ai-infra

免费

Infraestrutura para IA — GPU, servidores MCP, gateway de LLM, pipeline de inferência, custo de token, cache semântico e observabilidade de modelo. Use quando o usuário disser "GPU", "MCP server", "LLM", "inferência", "modelo", "gateway de IA", "custo de token", "rate limit da API de IA", "self-host de modelo".

中文简介正在准备中。当前显示原文说明。

pwdev-solucoes/pwdev-claude-marketplace32026年10月15日 更新

alt-text

免费

Escreve texto alternativo e descrição de imagem para peças de rede social. Use quando o usuário disser "alt text", "texto alternativo", "descrição da imagem", "acessibilidade da peça", "descrever para leitor de tela", ou automaticamente antes de qualquer publicação. Obrigatório em todas as peças — em carrossel, um por slide.

中文简介正在准备中。当前显示原文说明。

pwdev-solucoes/pwdev-claude-marketplace32026年10月15日 更新

Organiza, compara e seleciona entre as variações geradas, e publica a grade de comparação como Artifact para aprovação do time. Use quando o usuário disser "qual dessas", "comparar as variações", "escolher a melhor", "mostrar as opções", "montar grade de aprovação", ou após qualquer lote de geração. Selecionar é trabalho, não detalhe.

中文简介正在准备中。当前显示原文说明。

pwdev-solucoes/pwdev-claude-marketplace32026年10月15日 更新

Aumenta resolução e qualidade de imagem via Freepik/Magnific, ou orienta o processo manual quando não há chave. Use quando o usuário disser "upscale", "melhorar a resolução", "a imagem está pixelada", "Magnific", "Freepik", "aumentar a imagem", "imagem pequena demais". Gasta crédito — exige confirmação.

中文简介正在准备中。当前显示原文说明。

pwdev-solucoes/pwdev-claude-marketplace32026年10月15日 更新

CI/CD e automação — GitHub Actions, Terraform, OpenTofu, Ansible, n8n. Use quando o usuário disser "pipeline", "CI/CD", "GitHub Actions", "workflow", "terraform", "ansible", "n8n", "automatizar", "deploy automático", "IaC".

中文简介正在准备中。当前显示原文说明。

pwdev-solucoes/pwdev-claude-marketplace32026年10月15日 更新

查看 pwdev-solucoes 的全部技能

报告此技能的问题