translating-web
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data (HTML, Markdown, and web templates), which creates a potential surface for indirect prompt injection attacks where malicious instructions could be hidden in the source material.
- Ingestion points: The skill instructions in SKILL.md specify processing 'HTML, Markdown, templates, embedded data, and generated regions'.
- Boundary markers: The skill mandates keeping 'tags and field boundaries around the returned text' and identifying 'protected spans' to ensure translatable content is isolated from structural elements.
- Capability inventory: The skill primarily performs data classification and directs content to other internal components (translating-core, reviewing-translations); no privileged shell or file system operations are requested in the frontmatter.
- Sanitization: The instructions emphasize 'byte-for-byte' preservation of technical syntax and include a comprehensive 'QA Handoff' process to verify that structural parity and intent are maintained.
Audit Metadata