The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local Python scripts (scripts/preflight_check.py, scripts/flk_search.py, scripts/validate_output.py) to automate legal database searches, verify network connectivity, and validate the final output format against established 'Iron Rules'. The test suite also utilizes subprocess.run to verify script functionality locally.
[INDIRECT_PROMPT_INJECTION]: The skill processes legal metadata and text retrieved from external sources like the National Laws and Regulations Database and Northlake Law (Pkulaw). It mitigates risks associated with untrusted data through a mandatory evidence chain:
Ingestion points: External API responses from flk.npc.gov.cn and MCP-derived content.
Boundary markers: The skill enforces the use of specific markdown templates and explicit instructions to cross-reference multiple official sources.
Capability inventory: Uses local script execution for specific search and validation tasks.
Sanitization: Implements HTML tag stripping in flk_search.py and a domain blacklist in validate_output.py to prevent referencing untrusted third-party legal forums.
[EXTERNAL_DOWNLOADS]: The skill establishes network connections to official Chinese government domains (flk.npc.gov.cn, www.gov.cn) to retrieve legal documents and verify validity. These connections are performed via urllib.request within scripts to fetch authoritative metadata.
[SAFE]: No evidence of credential exfiltration, obfuscation, or persistence mechanisms was found. The skill's design focuses on reliability and authoritative sourcing for legal research.

lvdian-legal-search