The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process a wide variety of untrusted external materials, including case summaries, contracts, evidence catalogs, and chat logs. While the instructions include robust principles for facts/evidence extraction and data anonymization, the processing of untrusted content for logical reasoning and integration with external tools (such as legal search engines) establishes a vulnerability surface for indirect prompt injection.
Ingestion points: As defined in SKILL.md, the skill accepts inputs like client statements, contracts, payment vouchers, and record transcripts.
Boundary markers: The skill incorporates instructions to differentiate between evidence and statements, and to explicitly label missing information as '待补充' (to be supplemented), which serves as a grounding mechanism.
Capability inventory: The skill is capable of invoking legal search tools (yuandian-law-search, zhihe-legal-research), utilizing OCR for document conversion, and generating detailed Markdown reports.
Sanitization: The skill contains detailed rules for the anonymization and masking of personally identifiable information (PII) including names, IDs, and financial accounts, as specified in references/anonymization-output.md to mitigate data exposure risks.

legal-case-analysis