The Agent Skills Directory

[SAFE]: The skill is a well-engineered tool for public data access that adheres to security best practices. It correctly handles file system and network operations using defensive patterns like atomic writes and strict origin validation.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from several public City of Raleigh sources, creating a potential surface for indirect injection.
Ingestion points: Multiple scripts ingest remote data from ArcGIS, RSS, and HTML scrapers, specifically in scripts/raleighlib/civic.py, scripts/raleighlib/meetings.py, scripts/raleighlib/rfd_reports.py, and scripts/raleighlib/public_safety_stats.py.
Boundary markers: Not explicitly used when presenting tool output to the agent.
Capability inventory: The skill can make network requests to allowlisted domains and write files to the local system.
Sanitization: Robust sanitization is present, including the removal of terminal control characters from HTML fragments and the neutralization of spreadsheet formula triggers in CSV exports.
[EXTERNAL_DOWNLOADS]: The skill downloads civic data from fixed, official domains. All network requests are validated against a strict host allowlist in scripts/raleighlib/core.py which enforces HTTPS and strips sensitive headers during redirects.
[COMMAND_EXECUTION]: Static analysis identified a subprocess.run call in the unit tests (tests/test_raleigh.py). This is used for local integration testing of the CLI and does not expose a command execution vulnerability to the agent during normal skill usage.

raleigh