π¦ ClawHub
AANA Guardrail Skill
by @mindbomber
Monitors agent actions for safety and compliance, preventing violations, private data leaks, or irreversible steps by enforcing predefined decision rules.
TERMINAL
clawhub install aanaπ About This Skill
AANA Guardrail Skill For OpenClaw-Style Agents
Use this skill when a user asks the agent to produce or execute an action that may violate hard constraints, depend on missing evidence, expose private information, or require a safe correction path.
When To Call AANA
Call AANA before:
Command
Create an event JSON file and run:
python scripts/aana_cli.py agent-check --event path/to/agent_event.json
Event Shape
{
"event_version": "0.1",
"event_id": "unique-id",
"agent": "openclaw",
"adapter_id": "support_reply",
"user_request": "The user's request",
"candidate_action": "The answer or action the agent is about to take",
"available_evidence": ["Only facts actually available to the agent"],
"allowed_actions": ["accept", "revise", "ask", "defer", "refuse"]
}
Decision Rule
gate_decision is pass and recommended_action is accept, proceed.recommended_action is revise, use safe_response or revise the plan before acting.recommended_action is ask, ask the user for the missing information.recommended_action is defer, route to a stronger tool, human review, or verified system.recommended_action is refuse, do not execute the candidate action.Default Adapter Mapping
support_replytravel_planningmeal_planningresearch_summaryIf no adapter fits, scaffold one:
python scripts/aana_cli.py scaffold "new workflow name"