translating-documentation
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and parse untrusted external documentation (guides, API references, READMEs) which may contain malicious instructions intended to influence the agent's behavior during the translation or verification process.
- Ingestion points: External files are parsed during the 'Input Inspection' phase, including prose, code fences, and example data.
- Boundary markers: The instructions emphasize preserving document topology and distinguishing between 'explanatory prose' and 'non-translatable' technical elements like code fences.
- Capability inventory: The agent is instructed to 'parse, compile, or run' examples found in the documentation during the 'QA Handoff' phase.
- Sanitization: The skill relies on syntax validation (JSON/YAML/XML) and document topology parity checks to identify corrupted content, but does not explicitly filter for malicious prompts.
- [DYNAMIC_EXECUTION]: The 'QA Handoff' section contains instructions to ensure 'examples parse, compile, or run using the project's safe documented test method'. This pattern involves the execution of code snippets extracted from potentially untrusted documentation files at runtime.
Audit Metadata