Lightweight governance for multi-agent teams.
┌──────────────────────────────┐
│ MAIN SYSTEM │
│ (ClawStation / Orchestrator) │
└──────────────┬───────────────┘
│
▼
┌──────────────────────────────┐
│ DRILL SERGEANT SKILL │
│ (Agent Discipline & Task │
│ Enforcement) │
└──────────────┬───────────────┘
│
┌───────────┼───────────┐
▼ ▼ ▼
┌───────────┐ ┌───────────┐ ┌───────────┐
│ Task │ │ Behavior │ │ Monitor │
│ Manager │ │ Enforcer │ │ System │
│ Assigns │ │ Prevents │ │ Watches │
│ tickets │ │ random │ │ loops/ │
│ and jobs │ │ actions │ │ stale work│
└─────┬─────┘ └─────┬─────┘ └─────┬─────┘
│ │ │
▼ ▼ ▼
┌───────┐ ┌───────┐ ┌───────┐
│Agent 1│ │Agent 2│ │Agent 3│
│Coder │ │Tester │ │Ops │
└───────┘ └───────┘ └───────┘
Drill Sergeant helps enforce communication discipline and execution hygiene in shared work channels by identifying common operational failure patterns and producing clear corrective actions.
- Detect repeated/looping messages
- Flag routing violations (wrong channel / wrong audience)
- Surface missing ownership signals
- Highlight stale in-review work
- Reduce noisy status chatter
- Produce concise manager-ready summaries
- Define scope and rules for the review window
- Collect message/task signals from approved sources
- Detect violations using the checklist
- Deduplicate repeated findings
- Output owner-specific corrective actions
- Escalate only high-severity or repeated issues
Each finding should include:
typeseverity(low|medium|high)evidenceactionowner
Then summarize:
- Top 3 immediate actions
- Escalations (if any)
- All clear (when no actionable issues)
references/enforcement-checklist.mdreferences/message-templates.mdreferences/public-safety-checklist.md
This package is designed to be publish-safe:
- No personal identifiers
- No webhook/callback dependencies
- No API keys, tokens, or secrets
- No internal host/IP references
clawhub install drill-sergeantCopy this folder into your OpenClaw skills directory and load it per your normal skill workflow.
Drill Sergeant is the lightweight discipline layer.
For full multi-agent operations (orchestration, dashboards, automation workflows, and team command center), pair it with ClawStation: