The Agent Skills Directory

[SAFE]: The skill includes a dedicated security.py module that implements validate_project_root to prevent path traversal attacks by ensuring all file reads and writes are confined to the current project directory.
[SAFE]: The skill implements DoS protection in generator/security.py through the read_json_safe function, which enforces a 10MB file size limit for JSON parsing to prevent resource exhaustion attacks.
[SAFE]: The implementation avoids complex template engines, using Python's standard .format() method for test generation, which minimizes the risk of template injection and keeps dependencies restricted to the standard library.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves analyzing project source code and configuration files to generate test logic, creating an inherent ingestion surface for untrusted data. This is a characteristic of code-generation tools and is mitigated by the skill's restrictive file-writing and lack of immediate execution of generated content.
Ingestion points: generator/stack_detector.py and generator/app_type_detector.py perform recursive file searches and reads across the project root.
Boundary markers: Not utilized within the generated test templates to separate static logic from project-derived data.
Capability inventory: The skill possesses local file-writing capabilities via generator/utils.py and directory creation via ensure_directory.
Sanitization: Provided via path validation and shell character sanitization in the security.py module.

e2e-outside-in-test-generator