architect-review-live-calls
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted transcript content, serving as a vector for indirect prompt injection.
- Ingestion points: Conversation summaries and full transcripts are fetched from the ElevenLabs API via the conversations endpoints as documented in SKILL.md.
- Boundary markers: The instructions lack specific boundary markers or safety directives to the agent regarding the handling of potentially malicious content within transcripts.
- Capability inventory: Although this skill is restricted to read-only operations (GET), it is designed to route findings to a suite of sibling skills (e.g., architect-edit-workflows, architect-manage-procedures) that possess write capabilities for altering agent behavior.
- Sanitization: While PII redaction is required for reporting, there are no instructions to sanitize or validate the content of the transcripts themselves before the agent evaluates them for root causes.
- [SAFE]: The skill performs legitimate network requests to ElevenLabs' official infrastructure (api.elevenlabs.io) and its regional variants. These operations are aligned with the skill's monitoring purpose and do not interact with unauthorized or untrusted domains.
Audit Metadata