The Agent Skills Directory

[SAFE]: The primary ingestion logic in src/ingest.py is restricted to programmatic JSON transformations. It does not initiate network connections, access sensitive files, or execute system commands.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted alert data from external sources. Ingestion points: src/ingest.py reads data through iter_raw_alerts from sys.stdin or a file input. Boundary markers: Absent. Capability inventory: The skill is limited to file I/O and dictionary manipulation; it lacks network access, subprocess execution, or dangerous code evaluation functions. Sanitization: Data is parsed with json.loads and structural validation is performed in validate_alert.
[DYNAMIC_EXECUTION]: Testing components in tests/conftest.py and tests/test_ingest.py use dynamic module loading and sys.path manipulation to manage dependencies in the test environment. This is standard practice for skill isolation and does not affect the production logic.

ingest-azure-defender-for-cloud-ocsf