Listen
Pass
Audited by Gen Agent Trust Hub on Jun 20, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface because it ingests and stores potentially untrusted transcription data into its operational state.
- Ingestion points: Transcribed text and user corrections are captured in the 'Corrections', 'Patterns', and 'Context' sections of SKILL.md.
- Boundary markers: There are no explicit delimiters or clear separation instructions used to prevent the agent from interpreting embedded commands within the transcribed text.
- Capability inventory: The skill is designed to modify Speech-to-Text configurations using the config.patch tool, as described in config.md.
- Sanitization: The skill instructions do not specify any validation or sanitization steps for the ingested transcription content.
- [SAFE]: The configuration templates in config.md use standard placeholders (gsk_... and sk-...) for API keys rather than actual sensitive credentials, following secure documentation practices.
Audit Metadata