The Agent Skills Directory

[COMMAND_EXECUTION]: Internal validation and test scripts (validation/smoke.py and tests/test_universe_auditor.py) utilize subprocess.run to execute the project's own analysis script for functional testing.
Evidence: The calls use sys.executable to run the local script in a controlled manner during the build/test phase.
[SAFE]: The skill implements a modular architecture for data acquisition, delegating network operations to a designated sibling skill (skill-pandadata-api).
Evidence: SKILL.md and references/pandadata-integration.md define a clear boundary where the auditor skill remains offline and deterministic, reducing the risk of unauthorized network activity.
[SAFE]: Data processing in scripts/audit_universe.py includes strict schema validation for the input CSV files.
Evidence: The load_rows and analyze functions verify required columns (symbol, date, listed_at), enforce binary eligibility flags, and validate numerical return formats, which prevents the processing of malformed or malicious data structures.

survivorship-universe-auditor