fixing-metadata
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill consists entirely of instructional markdown providing best practices for metadata management and does not include any executable scripts or remote dependencies.
- [INDIRECT_PROMPT_INJECTION]: The skill handles metadata which may contain untrusted data, but it is considered safe as it explicitly incorporates defensive instructions.
- Ingestion points: Metadata fields within HTML or source code files (SKILL.md).
- Boundary markers: None explicitly defined as delimiters, but the skill advises the agent to treat dynamic data with caution.
- Capability inventory: File system modification (to fix metadata) and browser access (to verify social cards).
- Sanitization: The instructions explicitly mandate the escaping and sanitization of any user-generated or dynamic strings.
Audit Metadata