lov-image-translation-errata
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONPERSISTENCECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes text from user-supplied images, which could potentially contain malicious instructions aimed at the agent.
- Ingestion points: Image files, screenshots, and posters provided for translation review in SKILL.md.
- Boundary markers: The skill includes specific instructions in SKILL.md to treat all text extracted from images as quoted content rather than active instructions.
- Capability inventory: Vision-capable runtime, image editing tools, and file system access for profile storage.
- Sanitization: Mitigated via prompt-based instructions to isolate data from commands.
- [PERSISTENCE]: The scripts/profile_store.py script enables the persistence of user preferences, brand information, and skill-specific records across sessions by writing to local JSON files (profile.json).
- [COMMAND_EXECUTION]: The skill provides utility scripts (scripts/profile_store.py and scripts/validate_skill.py) that perform file system operations and validation, intended for execution within the agent environment.
- [EXTERNAL_DOWNLOADS]: The documentation describes installation using npx, which involves fetching the skill package from an external vendor registry.
Audit Metadata