The Agent Skills Directory

[DATA_EXFILTRATION]: The skill processes and stores sensitive litigation data (case numbers, parties, and court documents) into a local archive/ directory and case-specific folders. This behavior is central to its stated purpose as a legal management tool and includes explicit transparency warnings in the documentation regarding local data storage.
[INDIRECT_PROMPT_INJECTION]: The skill operates on an attack surface that involves ingesting untrusted data from SMS messages and downloaded PDF documents.
Ingestion points: SMS text input provided by the user and content parsed from court PDF documents (SKILL.md).
Boundary markers: No specific boundary markers or safety delimiters are defined in the instructions for the agent when processing these external data sources.
Capability inventory: The skill possesses capabilities to write to the local filesystem and perform network requests via curl and Playwright (SKILL.md, references/sms-patterns.json).
Sanitization: There are no explicit sanitization or filtering protocols mentioned for data extracted from these sources before it is used in file naming or reporting.
[EXTERNAL_DOWNLOADS]: The skill downloads legal documents from official Chinese judicial platforms, including zxfw.court.gov.cn, sd.gdems.com, and dzsd.hbfy.gov.cn. These references are consistent with the skill's documented functionality.
[COMMAND_EXECUTION]: The skill executes a local Python script (scripts/check_service_evidence.py) to validate service evidence data. Additionally, its test suite (scripts/test_service_evidence.py) utilizes subprocess.run to verify script exit codes during development.

court-sms