Agent Tool Risk

Use this skill for MCP, plugins, hooks, sub-agents, tool permissions, prompt flows, automation loops, and model-visible logs.

Workflow

1. Inventory the agent/tool surface: trigger, command, permissions, inputs, outputs, and persistence.
2. Classify risk:
   • prompt injection or untrusted content
   • excessive agency or destructive authority
   • secret exposure
   • data exfiltration through logs or MCP
   • command injection
   • stale or misleading model-visible memory
3. Confirm high-risk operations have human approval or guard hooks.
4. Ensure hidden logs are not loaded as default model context.
5. Ensure model-visible logs contain only durable, necessary retry context.
6. Update docs/harness/AGENT_SECURITY.md, docs/harness/SECURITY_POLICY.md, and docs/harness/SUBAGENT_PROTOCOL.md if behavior changes.