The Agent Skills Directory

[SAFE]: The skill operates within its defined scope of legal document management. It loads a configuration file from a specific local path (~/.claude/plugins/config/claude-for-legal-zh/employment-legal/CLAUDE.md) to establish context and procedural rules.
[DATA_EXFILTRATION]: While the skill manages sensitive legal information (complaints, evidence lists, and witness logs), it does not include any instructions for network transmission, exfiltration, or unauthorized data exposure. The data is handled locally through the creation of log files.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an ingestion point for user-provided data, which constitutes a minor attack surface.
- Ingestion points: The argument-hint field in SKILL.md captures a brief description of allegations from the user.
- Boundary markers: The instructions do not define specific delimiters to isolate user input from the rest of the prompt logic.
- Capability inventory: The skill includes file-writing capabilities (creating and updating logs) but no high-risk capabilities like network access or code execution.
- Sanitization: No specific sanitization methods for the user input are documented, but the processing is limited to narrative reporting rather than executable logic.

investigation-open