acestep
Pass
Audited by Gen Agent Trust Hub on Nov 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill centers on the execution of a local script,
tools/music_gen.py, via the shell. The provided examples demonstrate safe practices, such as using quoted here-documents (cat > /tmp/lyrics.txt << 'LYRICS') to safely manage multi-line text input without unintended shell expansion.\n- [CREDENTIALS_UNSAFE]: The skill correctly identifies the need for sensitive credentials, such as theRUNPOD_API_KEY, and explicitly instructs the user to store them in a.envfile, adhering to standard security practices for secret management.\n- [INDIRECT_PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection via user-supplied prompts and lyrics. Evidence Chain: 1. Ingestion points:--promptand--lyricscommand-line arguments. 2. Boundary markers: None specified in the shell instructions, though structural tags like[Chorus]are used in lyrics. 3. Capability inventory: Execution oftools/music_gen.pyvia subprocess and file creation in/tmp. 4. Sanitization: Not explicitly defined in the markdown instructions. The risk is considered safe because the primary output is audio, which does not provide a textual feedback loop to the agent.
Audit Metadata