con un clic
agile-v-evidence
Evidence collection and update behavior for Agile-V task lifecycle.
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Menú
Evidence collection and update behavior for Agile-V task lifecycle.
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Implementation behavior for builder sessions following the Agile-V process.
Apply Agentic Agile-V process control, task briefs, risk-based evidence gates, and human approval rules.
Risk level classification guidance (L0-L4) for Agile-V tasks.
Independent review behavior for verifier sessions in the Agile-V process.
Basado en la clasificación ocupacional SOC
| name | agile-v-evidence |
| description | Evidence collection and update behavior for Agile-V task lifecycle. |
Maintain an accurate, real-time evidence bundle throughout the task lifecycle. Evidence is collected from Git and CI — not from agent memory.
Collect evidence before claiming the task is done. Do not write "tests pass" until you have run them and recorded the output. Do not write "lint passes" until you have run the lint command.
.agentic-agile-v/tasks/AAV-XXXX/evidence_bundle.json
{
"task_id": "AAV-XXXX",
"risk_level": "L1|L2|L3|L4",
"changed_files": [],
"tests": {
"added": [],
"modified": [],
"run": [],
"results": []
},
"checks": {
"lint": {},
"typecheck": {},
"build": {}
}
}
git diff --name-only HEAD or git statusNever paraphrase CI or test output. Paste the relevant lines verbatim into the evidence bundle.
If using OpenHands, also add:
{
"agent_execution": {
"engine": "openhands",
"mode": "builder|verifier",
"session_id": "...",
"tool_log_path": "logs/openhands_tool_log.jsonl"
},
"scope_control": {
"allowed_paths": [],
"blocked_paths": [],
"changed_files_within_scope": true
}
}
changed_filestests.run and tests.resultschecksallowed_paths"pass" for a test you have not run"no issues" for a lint check you have not run"skipped" and the rationale — not "pass"