The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/run.py script uses subprocess.run to execute Docker commands, facilitating bug reproduction in an isolated environment. The Docker execution is hardened with flags like --cap-drop ALL, --security-opt no-new-privileges, and --network none.
[INDIRECT_PROMPT_INJECTION]: The skill processes external transcript data and issue descriptions. It includes explicit instructions to treat this content as evidence rather than instructions, and reproduction scripts are executed within isolated Docker containers to mitigate potential injection risks.
[DATA_EXFILTRATION]: The skill mandates the redaction of sensitive identifiers in generated artifacts and enforces network isolation during execution via Docker's --network none flag, preventing unauthorized data transmission.
[DYNAMIC_EXECUTION]: The scripts/test_validate.py script uses importlib to dynamically load and execute scripts/validate.py for testing purposes. This is a standard practice for unit testing and does not involve untrusted external code.

deck-repro