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
Risk Level
SAFE
Analyzed
Aug 13, 2026, 04:30 PM
Security Audit — agent-trust-hub — music-caption-rewriter