| name | admina-governance |
| description | Route all agent actions through Admina governance proxy for PII redaction, firewall enforcement, loop breaking, and forensic audit logging. Required for EU AI Act compliance.
|
| metadata | {"openclaw":{"emoji":"🛡️","requires":{"bins":["docker"],"env":["ADMINA_PROXY_URL"]},"install":[{"id":"docker","kind":"download","url":"https://admina.org/install.sh","label":"Install Admina sidecar"}],"homepage":"https://admina.org","primaryEnv":"ADMINA_PROXY_URL"}} |
Admina governance
When to activate
This skill is ALWAYS active. Every action you take must be
validated through Admina before execution.
How it works
Before executing any tool call, HTTP request, or shell command:
-
Send the action payload to the Admina proxy for validation:
curl -s -X POST ${ADMINA_PROXY_URL}/api/v1/validate \
-H "Content-Type: application/json" \
-d '{"action": "<ACTION_TYPE>", "payload": "<CONTENT>"}'
-
Check the response:
"allowed": true -> proceed with the action
"allowed": false -> report the block reason to the user
"modified": true -> use the modified payload (PII redacted)
-
After execution, log the result:
curl -s -X POST ${ADMINA_PROXY_URL}/api/v1/audit \
-H "Content-Type: application/json" \
-d '{"action": "<ACTION_TYPE>", "result": "<OUTCOME>"}'
Action types
llm_call -- any LLM API call
shell_exec -- any shell command
file_write -- any file creation/modification
http_request -- any outbound HTTP call
message_send -- any message to external platforms
If Admina proxy is unreachable
STOP. Do not proceed. Inform the user:
"Admina governance proxy is not responding. Cannot execute
ungoverned actions. Please check the sidecar status."
Dashboard
${ADMINA_PROXY_URL}/dashboard