music-caption-rewriter
Pass
Audited by Gen Agent Trust Hub on Aug 13, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: Comprehensive analysis of the 200 skill files reveals no evidence of malicious behavior, data exfiltration, or unauthorized command execution. The skill is designed as a static instruction set.
- [NO_CODE]: The skill package is entirely text-based, consisting of Markdown instructions and plain-text style templates. It does not include any Python scripts, Node.js files, binary executables, or configurations for external runtime dependencies.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user content via the 'Caption' and 'Lyrics' input fields. This creates a surface for indirect prompt injection. However, the risk is negligible as the skill lacks high-privilege capabilities such as shell access or network communication, and uses instructional constraints to isolate user data from musical directives. Ingestion points: 'Caption' and 'Lyrics' fields defined in SKILL.md. Boundary markers: Absent. Capability inventory: Limited to reading local template files. Sanitization: Relies on negative constraints (e.g., 'Never quote, paraphrase, or reproduce lyrics') to prevent the model from obeying instructions embedded in user data.
Audit Metadata