fathom
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes
subprocessto call external tools includingffmpeg,ffprobe,npm, andpython3for video processing and transcript analysis. These calls use argument lists rather than shell strings, which mitigates the risk of shell-based command injection. - [INDIRECT_PROMPT_INJECTION]: The skill retrieves meeting transcripts and AI-generated summaries from the Fathom API and saves them into markdown files intended for the agent's consumption. Malicious instructions embedded in these external transcripts could attempt to manipulate the agent's behavior during subsequent interactions.
- Ingestion points: Fathom API endpoints for meeting recordings, summaries, and transcripts accessed via
scripts/utils.py. - Boundary markers: The generated markdown files do not include explicit delimiters or instructions to the agent to disregard potential instructions within the fetched transcript data.
- Capability inventory: The skill possesses file-writing capabilities and triggers secondary processing through local script execution (in
scripts/fetch.py). - Sanitization: Data retrieved from the Fathom API is directly interpolated into markdown templates without sanitization or instruction-filtering.
Audit Metadata