voice
Pass
Audited by Gen Agent Trust Hub on Oct 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text and audio data for synthesis and cloning, which is passed to backend tools (submit_voice, submit_sound) without explicit sanitization or boundary markers defined in the prompt construction logic.\n
- Ingestion points: User-supplied text in submit_voice and sound descriptions in submit_sound parameters in SKILL.md.\n
- Boundary markers: Absent; instructions do not require delimiters or warnings to ignore embedded instructions in the synthesized content.\n
- Capability inventory: submit_voice, submit_sound, edit_item (timeline modification), pull_asset, and push_asset.\n
- Sanitization: Absent; no mention of escaping or validating the content of the text being synthesized.\n- [COMMAND_EXECUTION]: The instructions direct the agent to use the ffmpeg command-line tool to extract audio ranges from video assets. This operation presents a surface for command injection if the agent constructs shell commands using unsanitized file paths, asset IDs, or time ranges.\n
- Evidence: The instruction to 'extract the chosen range as supported audio with ffmpeg' is located in the reference audio resolution section of SKILL.md.
Audit Metadata