iblai-api-agent-safety

Mirrored from iblai/vibe · skills/agents/iblai-api-agent-safety/SKILL.md. This page is generated — edit it in the repository, not here.

Configure an agent's moderation and safety systems via the API: the moderation and safety enable flags, system prompts, and responses (all saved through the single settings/ endpoint), plus reviewing and deleting flagged-prompt moderation logs. Use when setting guardrails or auditing flagged content. The UI twin is /iblai-vibe-agent-safety.

Auth & conventions

• Base URL: https://api.iblai.app
• Header: Authorization: Api-Token $IBLAI_API_KEY on every request.
• Path vars: {org} = $IBLAI_ORG, {username} = $IBLAI_USERNAME, {mentor} = the agent's unique id (e.g. d17dc729-60fd-4363-81a0-f67d9318b03e).
• Settings writes go through one endpoint — PUT …/users/{username}/mentors/{mentor}/settings/ (JSON, multipart/form-data or form-encoded) — sending only the changed field(s).
• Not connected yet? Run /iblai-api-login first to populate IBLAI_ORG, IBLAI_USERNAME, and IBLAI_API_KEY.

Reads

• GET https://api.iblai.app/dm/api/ai-mentor/orgs/{org}/users/{username}/mentors/{mentor}/settings/ — load moderation/safety prompts, responses, and enable flags.
• GET https://api.iblai.app/dm/api/ai-mentor/orgs/{org}/users/{username}/moderation-logs/?mentor={mentor}&page={n}&page_size={n}&search={q}&start_time={iso}&end_time={iso}&ordering={field} — list the user messages the Moderation Prompt stopped. It only ever returns Moderation System rows. Rows carry id, prompt, reason (why it was flagged), target_system, date_created and mentor (the agent's numeric id). username={u} filters to the user who was flagged, not the caller. Each row names who was flagged with username, email and user_full_name; the last two are null for a row whose user has since been deleted, which a log row is expected to outlive.
• GET https://api.iblai.app/dm/api/ai-mentor/orgs/{org}/users/{username}/safety-logs/?mentor={mentor}&page={n}&page_size={n} — the agent replies the Safety Prompt blocked (Safety System rows). Same query parameters and row shape as moderation-logs/; no SDK data-layer hook reads it yet.

Writes

• PUT …/users/{username}/mentors/{mentor}/settings/ — update moderation/safety fields (send only changed keys):

  {
    "enable_moderation": "boolean",
    "enable_safety_system": "boolean",
    "moderation_system_prompt": "string",
    "moderation_response": "string",
    "safety_system_prompt": "string",
    "safety_response": "string"
  }

• DELETE https://api.iblai.app/dm/api/ai-mentor/orgs/{org}/users/{username}/moderation-logs/{id}/ — delete a flagged log (no body); …/safety-logs/{id}/ deletes a safety log. Destructive — confirm with the user first.

Example

Enable the moderation system and set its system prompt and response (only the changed fields are sent):

curl -X PUT \
  "https://api.iblai.app/dm/api/ai-mentor/orgs/$IBLAI_ORG/users/$IBLAI_USERNAME/mentors/$MENTOR/settings/" \
  -H "Authorization: Api-Token $IBLAI_API_KEY" \
  -F "enable_moderation=true" \
  -F "moderation_system_prompt=Flag any request for medical, legal, or financial advice." \
  -F "moderation_response=I can't help with that. Please consult a licensed professional."

Notes

• A field left out of the PUT is left unchanged — never resend the whole object.
• Safety and moderation prompts/responses/flags all persist through the same settings/ endpoint as the rest of the agent's configuration.
• Each system has its own log: moderation-logs/ for Moderation System catches, safety-logs/ for Safety System catches. Narrow either with search, start_time, and end_time.
• Deleting a moderation log only removes the audit record — it does not change the agent's safety configuration.
• The moderation check runs on the user's message before the agent sees it; the safety check runs on the agent's reply. A message caught by moderation is answered with moderation_response, and logged in moderation-logs/ while the agent's save_flagged_prompts is on (the default).