Runs 17 verification agents (Pre-wave + Waves 1-10), clears pending markers. USE WHEN pre-commit or /verify invoked.
Runs verification agents and clears pending file markers.
/verify # Full verification pipeline
/verify --fix # Auto-fix issues after verification
PRE-WAVE → library-researcher (1 sequential)
WAVE 1 → scanner-bpe + scanner-security + scanner-hallucination (3 parallel per batch)
WAVE 1B → wave1-judge (1 sequential)
WAVE 2 → code-reviewer + test-generator (2 parallel)
WAVE 3 → N parallel fixers (dynamic, file-level partitioned)
WAVE 3B → fix-validator (1 sequential)
WAVE 4 → integration-tracer + async-safety-auditor (2 parallel)
WAVE 5 → semantic-correctness-auditor + circular-import-detector (2 parallel)
WAVE 6 → import-resolver + dependency-scanner (2 parallel)
WAVE 7 → config-validator (1 agent)
WAVE 8 → hex-arch-remediator (1 agent)
WAVE 9 → smoke-test-runner (1 agent)
WAVE 10 → wiring-auditor + stub-detector + config-cross-checker (3 parallel, last)
human → team-lead → orchestrator → agents
TIMESTAMP=$(date +%Y%m%d-%H%M%S)
TEAM_NAME="siopv-verify-${TIMESTAMP}"
VERIFY_DIR="/Users/bruno/siopv/.verify-$(date +%d-%m-%Y)"
mkdir -p "$VERIFY_DIR"/{context7-cache,scans,fixes,reports,handoffs}
TeamCreate(team_name="${TEAM_NAME}", description="SIOPV /verify run ${TIMESTAMP}", agent_type="general-purpose")
Agent(
description="Orchestrate /verify pipeline",
prompt="Read and follow the orchestrator protocol at ${CLAUDE_SKILL_DIR}/orchestrator-protocol.md — read the complete file before taking any action. Also read ${CLAUDE_SKILL_DIR}/wave-prompts.md for all agent prompt templates, ${CLAUDE_SKILL_DIR}/agent-rules.md for universal rules, and ${CLAUDE_SKILL_DIR}/thresholds.md for pass/fail criteria. Team name: ${TEAM_NAME}. Verify dir: ${VERIFY_DIR}. Report all spawn requests and wave results to team-lead.",
subagent_type="general-purpose",
team_name="${TEAM_NAME}",
name="orchestrator",
mode="acceptEdits",
run_in_background=True
)
You are "team-lead" in this team. The orchestrator addresses you as "team-lead".
MANDATORY: You MUST NOT spawn any agents without explicit human approval first.
Spawn requests: When orchestrator sends a SPAWN REQUEST:
a. STOP. Present the spawn plan to the human.
b. WAIT for human to explicitly approve.
c. Only after approval: spawn the requested agents with the exact specs provided.
d. Notify orchestrator: SendMessage(to="orchestrator", message="SPAWN CONFIRMED — Wave {N}: all agents spawned.")
Wave reports: When orchestrator delivers a WAVE {N} REPORT:
a. Present the full report to the human.
b. WAIT for explicit human approval before proceeding.
Approval relay: After human approves:
SendMessage(to="orchestrator", message="WAVE {N} APPROVED — proceed.")
Change requests: Relay human feedback to orchestrator. Wait for resolution.
Never fix issues yourself — always delegate to orchestrator.
Silence is not approval. If in doubt, ask the human.
TeamDelete()