architect-migrate-to-procedures
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes potentially untrusted data from an existing agent's system prompt during the migration process.
- Ingestion points: Step 2 in
SKILL.mddescribes reading theconversation_config.agent.prompt.promptfield from the ElevenLabs API. - Boundary markers: Absent. The instructions do not define delimiters or warnings to ignore instructions embedded within the prompt being analyzed.
- Capability inventory: The skill has write access to the agent's configuration, including the ability to create new procedures via
POST /v1/convai/agents/$AGENT_ID/branches/$BRANCH_ID/proceduresand update the system prompt viaPATCH /v1/convai/agents/$AGENT_ID. - Sanitization: Absent. The skill does not mention sanitizing or validating the extracted prompt content before using it to generate procedures or modify the agent configuration.
Audit Metadata