omnicaptions-translate
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted caption files which may contain malicious instructions designed to influence the LLM translation process.
- Ingestion points: Captions are read using the Read and Glob tools for processing.
- Boundary markers: While instructions specify format preservation, there are no explicit delimiters to isolate untrusted caption text.
- Capability inventory: The skill possesses Write capabilities and uses Bash to execute the omnicaptions CLI.
- Sanitization: No sanitization of the input caption text is implemented.
- [EXTERNAL_DOWNLOADS]: The skill installs a library from a vendor-controlled repository.
- Evidence: Installation command uses the vendor's GitHub Pages domain at https://lattifai.github.io/pypi/simple/.
- [COMMAND_EXECUTION]: The skill executes local commands through a specific CLI utility to perform translation tasks.
- Evidence: Multiple shell command examples using the omnicaptions tool are provided.
Audit Metadata