The Agent Skills Directory

[DATA_EXPOSURE]: The skill reads from local configuration and case management files located at ~/.claude/plugins/config/claude-for-legal/ and within matter workspaces. This access is limited to its specific functional domain (employment law) and is used to retrieve jurisdiction-specific rules and matter context.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection as it processes untrusted offer letters provided by users. However, it implements robust mitigations, including strict source-citation requirements (e.g., [yuandian检索], [联网检索——需复核]), an instruction to never supplement missing research with model hallucination, and a mandatory 'Follow-up Action Threshold' that requires explicit confirmation of lawyer oversight before providing a final recommendation.

hiring-review