dictation
Pass
Audited by Gen Agent Trust Hub on Aug 6, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes user-supplied voice dictations which are treated as commands, creating an indirect prompt injection surface where instructions hidden in the transcript could be executed by the agent.\n
- Ingestion points: The skill accepts a
<dictated request>argument inSKILL.mdthat is directly processed by the assistant.\n - Boundary markers: The instructions lack explicit delimiters to isolate user-provided dictation from the agent's internal logic, though the mandatory
Heard:readback acts as a validation checkpoint.\n - Capability inventory: The agent is instructed to use tools for repository state inspection and read-only checks to resolve ambiguous transcription terms.\n
- Sanitization: The protocol includes a proactive safety measure requiring the assistant to replace secrets or sensitive values, such as API tokens, with role labels like
[API token]during the readback process.
Audit Metadata