review-llm-annotations-and-improve-prompt
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted human labels, notes, and transcripts that could contain instructions targeting the LLM judge or agent. Evidence: 1. Ingestion points: Human labels, reviewer notes, and transcripts are ingested as described in SKILL.md. 2. Boundary markers: Absent. 3. Capability inventory: The skill uses the coval CLI and makes network requests to the vendor API. 4. Sanitization: None mentioned.
- [COMMAND_EXECUTION]: The skill instructs the agent to use the coval CLI tool for testing metrics and retrieving conversation details, which is a standard functionality provided by the vendor.
- [EXTERNAL_DOWNLOADS]: The skill interacts with the vendor's official API to fetch metric details and labels.
Audit Metadata