The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it is designed to ingest and process untrusted external data from uploaded legal documents.
Ingestion points: SKILL.md (File Handling section) specifies processing uploaded PDF and DOCX files.
Boundary markers: While the skill requires structured outputs and safety watermarks, it lacks explicit instructions to ignore instructions embedded within the source documents.
Capability inventory: The agent utilizes pdf-reading, docx, and pdf skills to handle files.
Sanitization: No explicit filtering of source document content is performed.
[SAFE]: The skill includes extensive professional disclaimers, a 'Delegation Gate', and mandatory watermarking (⚠ DRAFT) to ensure users are aware that outputs require legal review.
[SAFE]: No evidence of malicious behavior, credential harvesting, or unauthorized network access was detected in the skill code or reference materials.

legal-translation