The Agent Skills Directory

[SAFE]: The provided Python script scripts/dpa_clause_check.py is a self-contained, deterministic tool. It uses only standard libraries (argparse, json, sys) to compare input flags against a static checklist of GDPR articles. It does not perform network operations, file system modifications, or dynamic code execution.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external text (contract contents provided by the user), creating a surface for potential instructions embedded in data.
Ingestion points: The skill instructions specify that the user will paste contract text for review ("wklejona treść umowy trafia do modelu").
Boundary markers: No specific delimiters or "ignore instructions" warnings are defined in the skill instructions for the ingested text.
Capability inventory: The skill is allowed to use Bash and Read tools, and executes a local Python script for clause validation.
Sanitization: No explicit sanitization or filtering of the ingested legal text is described.
Mitigation: The risk is mitigated by the requires-human-approval: true metadata, ensuring a human reviews the draft/redline output before use, and the data-residency: local constraint.

rodo-ropa-dpa-pl