The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill has a potential surface for indirect prompt injection because it ingests and processes content from local project files during compliance checks.
Ingestion points: The check_asi_01 function in SKILL.md uses pathlib.Path(project_path).rglob("*.py") to read the contents of external Python files.
Boundary markers: None are present; the skill processes raw file content directly for pattern matching.
Capability inventory: The skill performs local file read operations using read_text(). It does not perform network operations or file writes.
Sanitization: The skill does not sanitize content but uses it strictly for regex-based pattern detection to generate a report.
[EXTERNAL_DOWNLOADS]: The skill references and recommends the installation of the agent-governance-toolkit package from Microsoft's official GitHub repository.
[COMMAND_EXECUTION]: The documentation contains examples of unsafe command execution (e.g., subprocess.run(shell=True)) specifically to teach users what to avoid and to provide patterns for the auditing script to detect in target codebases.

agent-owasp-compliance