media-index

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and storing untrusted metadata within the 'items' argument. This data is saved to a local JSON file without sanitization for downstream consumption.
  • Ingestion points: The 'items' input field in SKILL.md captures metadata, tags, and descriptive strings.
  • Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present within the index schema for these fields.
  • Capability inventory: The skill utilizes 'shell', 'read', and 'write' tools to manage local files and execute the indexing script.
  • Sanitization: There is no documented validation or filtering of the user-supplied metadata strings before they are written to the index file.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 08:22 AM
Security Audit — agent-trust-hub — media-index