The Agent Skills Directory

[COMMAND_EXECUTION]: The test suite in tests/test_audit_context.py utilizes subprocess.run to verify the audit script's functionality. These calls are considered safe as they are used solely for regression testing within the local environment and are restricted to invoking the skill's own Python script.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external session logs and request dumps, which constitutes a surface for indirect prompt injection. However, the audit script mitigates this risk by performing strict type checking and structural validation on the JSON data before processing, and the results are presented as metrics rather than raw content.

context-audit