bilingual-translate

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied markdown files, which acts as a potential surface for indirect prompt injection.\n
  • Ingestion points: Markdown source files in data/markdown/ and wave selection via chapters.json.\n
  • Boundary markers: The skill uses <!-- TODO: 翻譯 --> placeholders to isolate translation tasks and strictly instructs the agent to preserve original blockquotes (>) byte-for-byte.\n
  • Capability inventory: File system access to write drafts and final translations, and execution of local Python scripts (scripts/*.py) and git commands.\n
  • Sanitization: Structural validation is performed via bilingual_prep.py and terminology preflight checks.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 06:03 PM
Security Audit — agent-trust-hub — bilingual-translate