architect-get-agent-ci-green
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill involves analyzing agent transcripts and rationale messages produced during test runs. These messages represent untrusted data that could potentially contain adversarial instructions designed to influence the analyzing agent.
- Ingestion points: Processes data from agent.json and test invocation results fetched via the ElevenLabs API.
- Boundary markers: No specific delimiters or safety warnings are implemented to separate transcript data from instructions.
- Capability inventory: The agent uses curl to perform read and write operations on test definitions via the API.
- Sanitization: No explicit sanitization of transcript content is performed before processing.
- [COMMAND_EXECUTION]: The skill uses curl commands to interact with the ElevenLabs API for running test suites, polling results, and updating test metadata. These are standard operations for the skill's purpose.
- [EXTERNAL_DOWNLOADS]: Fetches configuration and test data from the official ElevenLabs API domain (api.elevenlabs.io). These downloads are from a well-known service associated with the skill author.
Audit Metadata