The Agent Skills Directory

[COMMAND_EXECUTION]: Multiple Python scripts in the scripts/ directory perform local shell command execution to facilitate the agent workflow. check_environment.py and discover_runners.py utilize subprocess to run version checks for developer tools like Git, Node.js, and Python. panel_round.py and runner_jobs.py are used to execute repository-local runner scripts. review_evidence.py allows for the execution of project-specific tests (e.g., npm test) to capture verification results. These are standard operations for an engineering-focused tool and are executed within defined project boundaries.
[REMOTE_CODE_EXECUTION]: A reference to a remote script execution pattern (curl | sh) was flagged. In this context, it is a false positive because the pattern is part of the skill's security mechanism. hooks/dangerous-patterns.txt defines a regular expression to block such patterns from being executed by the agent, and hooks/test-guard.sh uses the pattern as a test case to ensure the guardrail is functioning correctly.
[EXTERNAL_DOWNLOADS]: The skill includes external URLs in model-routing.json and check_environment.py pointing to official documentation, model provider APIs, and tool installers. These are provided as informational references for the user or as manual setup guidance and are not automatically downloaded or executed by the skill's scripts.

shared