Install
openclaw skills install claw-guardirails-skillEnforces guarded execution with safe_exec, safe_send, and safe_action. Use when a task may run shell commands, send channel messages, or call external APIs/actions that can mutate data or state.
openclaw skills install claw-guardirails-skillApply this skill whenever a request can:
safe_exec instead of raw exec.safe_send instead of direct channel-post tools.safe_action for generic external/API side effects.require_approval, stop and wait for explicit approval flow.cwd for safe_exec.safe_action, include explicit resources where possible.safe_execUse for shell commands only after permission resolution.
Expected input shape:
{
"command": "git",
"args": ["status"],
"cwd": "/workspace/project"
}
safe_sendUse for outbound channel messages.
Expected input shape:
{
"channel": "telegram:ops-room",
"message": "Deployment done",
"channelType": "telegram"
}
safe_actionUse for side-effect actions that are not plain shell or plain chat send.
Expected input shape:
{
"action": "gmail.delete_message",
"payload": { "messageId": "..." },
"resources": [
{ "kind": "unknown", "value": "email-api", "operation": "delete" }
]
}
allow: continue and return runtime result.deny: return denial with reason code; do not retry with alternate dangerous tools.require_approval: surface approval id/reason and wait for /approve <id> or RPC approval.git status, rg, ls) before mutable ones.