The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/claims.py uses importlib.util to dynamically load and execute the local sibling script totality_lint.py. While the path is resolved relative to the script's location within the skill folder, dynamic module loading is a form of runtime code execution.
[INDIRECT_PROMPT_INJECTION]: The tool possesses an attack surface for indirect prompt injection by processing untrusted repository data.
Ingestion points: The inventory and scan functions in the script files read the content of all Python files in the target repository using pathlib.Path.read_text.
Boundary markers: The tool's output does not implement specific delimiters or warnings to isolate ingested content from the AI agent's instructions.
Capability inventory: The skill is limited to local AST analysis and reporting; it has no capabilities for file writing, network access, or subprocess execution.
Sanitization: No sanitization or escaping of the ingested source code or docstring content is performed before outputting the results.
[SAFE]: No network operations, hardcoded credentials, or privilege escalation patterns were detected. The tool relies exclusively on standard Python libraries and performs all operations locally.

declaring-invariants