The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/test_examples.py performs dynamic loading and execution of Python code. It utilizes importlib.util.spec_from_file_location and exec_module to load and run assets/examples/explicit_behavior.py using a path computed at runtime relative to the script's location.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and act upon untrusted external repositories, creating an attack surface for indirect prompt injection.\n
Ingestion points: Codebase content, repository conventions, and test files within the target project (referenced in SKILL.md procedure step 1 and 6).\n
Boundary markers: The skill contains instructions to respect user boundaries and stop before destructive actions or credential use, but these do not prevent the agent from being influenced by data within the code being reviewed.\n
Capability inventory: The agent is authorized to execute Python scripts (scripts/test_examples.py) and arbitrary project checks (linters, tests, build commands).\n
Sanitization: There is no evidence of input validation or sanitization for the code or test outputs ingested from the target repository.
[COMMAND_EXECUTION]: The procedure in SKILL.md (Step 6) directs the agent to 'Run the project's existing checks and a behavior-focused test.' This involves the execution of shell commands and scripts defined in the external repository environment, which could result in the execution of malicious code provided by an untrusted source.

apply-pep20-to-codebases