The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface through its ingestion of untrusted data files.
Ingestion points: The skill ingests user-provided business profiles via scripts/applicability_check.py and analyzes draft markdown memos in scripts/citation_audit.py.
Boundary markers: Absent. The skill does not explicitly define delimiters or instructions for the AI agent to disregard potential instructions embedded within user-supplied text.
Capability inventory: The skill maintains the ability to execute CLI commands (utilized in tests/) and write files to the local system (scripts/generate_docx_memo.js).
Sanitization: Input processing relies on deterministic JSON parsing and regular expressions, but raw text from memos is audited without explicit sanitization against embedded prompts.
[COMMAND_EXECUTION]: Python scripts within the unit test suite utilize subprocess.run to validate tool functionality.
Evidence: Implementation found in tests/test_applicability_check.py, tests/test_precedent_match.py, tests/test_citation_audit.py, and tests/test_conflict_resolver.py.
Context: These commands are strictly limited to invoking the skill's own CLI tools using local paths for validation and testing purposes.
[EXTERNAL_DOWNLOADS]: The skill implementation and documentation reference an external package for deliverable generation.
Evidence: README.md and scripts/generate_docx_memo.js reference the docx library.
Context: The skill utilizes the standard docx library from the official NPM registry to produce formatted Microsoft Word documents.

us-state-privacy-navigator