Sentinel — Veto & Enforcement Authority
Sentinel — Veto & Enforcement Authority
What it is: The highest-authority AI governance agent — monitors every other agent, enforces your policy, and vetoes dangerous actions before they execute.
Who it's for: Teams running 3+ autonomous AI agents who need a kill switch, audit trail, and real-time policy enforcement.
What it does:
- Monitors all agent activity across every workspace in real time
- Vetoes any action that violates your constitutional rules
- Detects credential exposure, unauthorized spawning, and policy breaches
- Maintains a full append-only audit log of every agent decision
- Executes incident response — stops agents, logs violations, alerts operator
What it won't do:
- Does not execute tasks itself — governance and enforcement only
- Does not modify governance rules without explicit operator authorization
- Does not share cross-workspace data with execution-tier agents
Needs:
- AGENTS.md governance file defining the rules Sentinel enforces
- OpenClaw gateway or compatible multi-agent runtime
- Read access to all agent workspaces
Produces:
- Real-time violation alerts
- Append-only audit log (JSONL)
- Incident reports with action taken
- Policy enforcement decisions (allow / veto / flag)
Install in 60 seconds: Extract ZIP, run installer. Full guide: https://hivedeck.ai/pages/how-to-install
Works with: OpenClaw, Claude Code, Cursor, AutoGPT, CrewAI, LangGraph
Sentinel operates at Tier 1 Governance — the final authority on what runs
The only agent that can stop all the others.
Sentinel is the highest-authority agent in the HiveDeck architecture. It watches every other agent, evaluates every action against your defined policy, and vetoes anything dangerous before it executes. File system monitoring, credential exposure detection, constitutional rule enforcement — Sentinel handles all of it.
One-command install. Runs continuously in the background. After the ClawHavoc community skills incident, this is what responsible autonomous deployment looks like — every agent action reviewed, every violation stopped, every decision logged.
What you actually get (12 files and folders, ready in under 60 seconds)
- SOUL.md — Sentinel's complete authority charter and enforcement logic
- install.sh / uninstall.sh / verify.sh — one-command setup, teardown, and health check
- skills/ — full governance and enforcement tool definitions
- TOOLS.md — every monitoring and veto tool, pre-configured
- AGENTS.md integration guide — how to wire Sentinel into your existing agent roster
- ClawHavoc safety guide — incident response protocols and boundary definitions
- config/ / scripts/ / docs/ — everything pre-configured, nothing to guess
Install in 4 steps
- Download the ZIP
cd ~/.openclaw/agents && unzip sentinel.zip./install.sh- Tell your Chief of Staff: "Sentinel has authority to veto any agent action"
Works with Claude 3.5/4 or GPT-5. Install Sentinel before any execution agents — it needs to be running first to establish the governance boundary everything else operates under.
What Sentinel enforces
- Real-time veto of dangerous agent actions before execution
- File system change monitoring across all agent workspaces
- Credential and secret exposure detection
- Constitutional rule enforcement — your policy, applied consistently
- Complete audit trail of every veto, every review, every decision
- Escalation to human oversight when policy is unclear
Tier 1 Governance. Not optional.
Running autonomous agents without an enforcement layer is the incident waiting to happen. Sentinel is what separates "AI assistant" from "AI system" — the difference between hoping your agents stay in bounds and knowing they will. Every action bounded. Every violation stopped. You stay in control.
Part of the Governance Stack ($149 — save $89 vs. buying separately). Install with Auditor and Shield for the complete Tier 1 authority layer.
New to HiveDeck? Step-by-step install guide →
Related agents: Shield — Defensive Hardening Agent | Auditor — Compliance & Validation Agent | Governance Stack
Install guide: hivedeck.ai/pages/how-to-install