architect-agent-memory
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest user-supplied 'facts' or 'corrections' and append them directly to the agent's system prompt (
conversation_config.agent.prompt.prompt) or knowledge base documents. - Ingestion points: User-provided facts or behavioral corrections processed at runtime (SKILL.md).
- Boundary markers: The instructions do not define delimiters or safety warnings for the injected text within the system prompt to distinguish user-provided 'facts' from original system instructions.
- Capability inventory: The skill utilizes
PATCHandPOSTmethods against thehttps://api.elevenlabs.io/v1/convaiendpoints to modify agent configuration and knowledge sources (SKILL.md). - Sanitization: The skill includes a manual 'sanity-check' step for the agent to verify if the correction is factual and scoped, but it lacks programmatic sanitization or escaping to prevent a user from providing a 'fact' that contains new system-level instructions or overrides.
Audit Metadata