meetings-debrief
Pass
Audited by Gen Agent Trust Hub on Jul 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local shell commands including a custom CLI tool 'minutes' and standard utilities such as ls, mkdir, and chmod. These are used to list recordings, check for prep files, and manage the relationship data directory with restricted permissions (0600).
- [PROMPT_INJECTION]: The skill has an indirect prompt injection surface through the processing of meeting transcripts. 1. Ingestion points: Reads transcripts and metadata from meeting recordings in Phase 2. 2. Boundary markers: No delimiters or instructions are used to isolate transcript content from agent instructions. 3. Capability inventory: The skill can execute subprocesses via the 'minutes' CLI and write files to the local filesystem. 4. Sanitization: There is no evidence of filtering or sanitizing transcript content before it is used to update persistent relationship files.
Audit Metadata