audit-transcription
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process auto-transcribed meeting notes and audio transcripts, which are untrusted external data sources that may contain malicious instructions designed to override the agent's behavior.
- Ingestion points: External transcription text shared by the user or read from files (SKILL.md workflow step 1).
- Boundary markers: The skill suggests using markdown callouts (
> [!warning]) to separate gist interpretations from raw transcript text, providing some visual and structural delimitation. - Capability inventory: The agent is authorized to read project notes, glossaries, and codebase files to establish context, and is instructed to perform in-place file edits after receiving user confirmation.
- Sanitization: There is no explicit sanitization of the input text mentioned, but the skill implements a high degree of human-in-the-loop control, requiring the agent to flag issues and wait for user approval before applying any corrections, which serves as a procedural mitigation against accidental execution of injected commands.
Audit Metadata