agentleFS
GustavoAlecioGustavoAlecio/claude-setup/skills/verify/SKILL.md

verify

Skill0 starsChanged 27 days agoMITView on GitHub

Etapa 5 do Fluxo Smart — G1 arquitetural do ciclo inteiro + G2 QA (critérios de aceite, testes, app rodando) via workflow smart-verify; reprovou → reentra na escada automaticamente

188 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use instinct-export, instinct-import and 9 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • TheKREATORSSkill

    instinct-export

    Export instincts from project/global scope to a file

    Also in 3 projects that use this

  • TheKREATORSSkill

    instinct-import

    Import instincts from file or URL into project/global scope

    Also in 3 projects that use this

  • AviadPSkill

    risk-assess

    Assess the risk and blast radius of a proposed code change before implementing it. Evaluates CI impact, framework-wide vs localized scope, rollback path, and identifies blindspots. Use when user says "what's the risk", "is this safe", "assess the impact", "risk assess", "blast radius", or before applying framework-wide or CI-affecting changes. Do NOT use for code quality review (use code-quality) or plan validation (use design-validator).

    Also in 3 projects that use this

  • diego-tobalinaSkill

    build-fix

    Fix build and compilation errors systematically. Run when build fails.

    Also in 6 projects that use this

  • AaravChadhaSkill

    secure

    Security review with a confidence gate - a finding exists only with a concrete exploit scenario and a high/medium/low confidence rating; everything else goes to a worth-hardening list, never inflated. Sweeps five surfaces: auth gates, secrets hygiene, injection and unsafe sinks, unsafe deserialization/crypto/transport, and LLM tool-use trust boundaries. Reports only, never fixes. Use when the user asks for a security review or to check vulnerabilities, secrets, or auth.

    Also in 3 projects that use this

  • Jurel89Skill

    ai-slop-cleaner

    Clean AI-generated code slop with a regression-safe, deletion-first workflow and optional reviewer-only mode

    Also in 4 projects that use this

  • +1 more from TheKREATORS

The file

Files
SKILL.md

1. Detectar projeto

PROJECT_PATH=$(git rev-parse --show-toplevel 2>/dev/null || pwd)
source ~/.claude/bin/get-project.sh
PROJECT_NAME=$(get-project-name)
WF_DIR="$HOME/.claude/workflow/$PROJECT_NAME"

2. Pré-requisitos

Leia $WF_DIR/current.json, spec.md, plan.md, tasks.md.

  • Tasks pendentes → "Há tasks pendentes. Rode /implement primeiro."
  • Sem exec.base_checkpoint → o ciclo não passou pelo /implement novo; pare e explique.

Mesmo profile de stack do /implement (seção 3 de lá).

3. Preparar run

bash ~/.claude/bin/capture-metrics.sh start verify "$PROJECT_NAME" "$PROJECT_PATH"
RUN_ID="verify-$(date -u +%Y%m%dT%H%M%SZ)"
RUN_DIR="$WF_DIR/runs/$RUN_ID"
mkdir -p "$RUN_DIR"
bash ~/.claude/bin/wf-checkpoint.sh create "$PROJECT_PATH" > "$RUN_DIR/revision-start.tree"

revision-start.tree é a árvore que o verify vai julgar. O persist compara com a árvore no fim: se alguém mexeu no repo durante o verify, o resultado cai para unverified (REVISION_CHANGED). Não edite o repo entre o disparo e o persist.

Relatório: início da etapa

python3 ~/.claude/bin/wf-report.py stage-start verify --workflow-dir "$WF_DIR" ${CLAUDE_FLOW_SESSION_ID:+--session "$CLAUDE_FLOW_SESSION_ID"} || true

Antes de disparar, audite o harness na fase verify (inclui o g2_device: plataforma existe no repo e device conectado):

python3 ~/.claude/bin/harness-audit.py --repo "$PROJECT_PATH" --stack <stack> --phase verify --quick --text
  • FAIL → não dispare: o G2 voltaria inconclusive e gastaria a rodada. Mostre os findings; device desconectado → peça para conectar e espere o OK; G2-PLATFORM → o g2_device do profile não serve para este repo, decisão do usuário. --fix só com OK.
  • major → mostre e siga.

4. Disparar

Ferramenta Workflow com name: "smart-verify" e args:

{
  "project_path": "<PROJECT_PATH>",
  "workflow_dir": "<WF_DIR>",
  "run_dir": "<RUN_DIR>",
  "stack_name": "flutter",
  "stack": { "...profile..." },
  "base_checkpoint": "<exec.base_checkpoint>",
  "checkpoint": "<exec.checkpoint>",
  "revision_start": "@<RUN_DIR>/revision-start.tree",
  "plan_path": "<WF_DIR>/plan.md",
  "spec_path": "<WF_DIR>/spec.md",
  "lessons_path": "<se existir>",
  "tasks": [ "...todos os tasks.items, com tier/tier0/attempts/files_changed/tests/affects..." ],
  "max_rounds": 2,
  "max_attempts": 5,
  "per_tier": 2
}

Informe em 1 linha e encerre o turno.

5. Ao receber o resultado

Grave em $RUN_DIR/workflow-result.json exatamente como veio (não edite status, reason nem decision) e persista e importe o run no relatório. O persist vincula o resultado à revisão e grava exec.verify no current.json; o status que vale é o do result.json persistido, não o que o workflow devolveu:

python3 ~/.claude/bin/wf-event.py persist --run-dir "$RUN_DIR" --workflow-dir "$WF_DIR" --result-file "$RUN_DIR/workflow-result.json"
python3 ~/.claude/bin/wf-report.py import-run verify --workflow-dir "$WF_DIR" --run-dir "$RUN_DIR" || true

Relatório:

## Verify — <status> · <decision.reason_code> (rodadas: N)

Critérios: X PASS · Y FAIL · Z PARTIAL · W UNTESTABLE · revisão <decision.revision_id[:12]>

### G1 — arquitetura
| Reviewer | Veredito | Bloqueantes |
|---|---|---|

### G2 — critérios de aceite
| # | Critério | Status | Evidência |
|---|---|---|---|

### Reentradas
- r1: T2, V1 → done (T2 sonnet→opus)

A decisão é do código (decide no smart-verify.js, fail-closed); você só relata. verified exige G1 todo pass, G2 pass, ao menos um critério PASS com evidência, nenhum FAIL/PARTIAL/UNTESTABLE e nenhum finding critical/major. Rode capture-metrics.sh end verify em qualquer status: ele grava verified só quando exec.verify.status é verified, senão unverified.

Por status do result.json persistido:

  • verified → marque os critérios PASS como [x] na spec.md. Auto on: siga para /complete. Auto off: "Verificado. Fechar o ciclo? /complete".
  • unverified → mostre decision.reasons e o que cada um pede: QA_SKIPPED/CRITERIA_UNTESTED/NO_POSITIVE_EVIDENCE → QA humano dos critérios listados; G2_INCONCLUSIVE/G1_INCONCLUSIVE → resolver ambiente (device, backend, credencial) e rodar de novo; G2_MALFORMED/NO_CRITERIA → o QA não devolveu critérios válidos, rode de novo; G2_FAIL/G1_FAIL/CRITERIA_FAILED sem findings → o gate reprovou sem dizer onde, rode de novo ou investigue; REVISION_CHANGED/REVISION_UNBOUND → o repo mudou durante o verify, rode de novo. Não marque nada na spec. Parada incondicional (AskUserQuestion, mesmo com auto on): rodar /verify de novo, ou seguir para /complete sabendo que ele só fecha com exceção humana.
  • blocked / backtrack → mesmo tratamento do /implement (diagnosis ToT, protocolo de desvio), sempre como AskUserQuestion (parada incondicional, mesmo com auto on). Nunca decida sozinho.

Findings minor/nit dos reviewers: liste no fim como "não bloqueantes", sem abrir task.

Relatório: fim da etapa. Escreva com Write um resumo de 3 a 10 linhas em $WF_DIR/.stage-summary.md (nunca interpole texto em shell) e rode (acrescente --status blocked em tudo que não for verified):

python3 ~/.claude/bin/wf-report.py stage-end verify --workflow-dir "$WF_DIR" --summary-file "$WF_DIR/.stage-summary.md" || true

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.