The Agent Skills Directory

[SAFE]: The skill facilitates the installation of verifier loops (format, build, test, and CI) through a local detection script and templates. It follows a structured workflow that prioritizes user confirmation before any modifications are made to the target repository.
[COMMAND_EXECUTION]: The evaluation and test suites (eval/run_eval.py, assets/test_detect_stack.py) utilize subprocess.run to execute the skill's own detection script for validation purposes. These executions are local, do not use shell interpolation for user input, and are standard for software testing suites.
[DATA_EXPOSURE_AND_EXFILTRATION]: The detection script (assets/detect_stack.py) reads local repository metadata and configuration files (such as package.json, Makefile, and go.mod) to generate installation plans. The script is designed for offline use, relies strictly on the Python standard library, and performs no network operations.
[INDIRECT_PROMPT_INJECTION]: While the skill ingests untrusted data from the repository being analyzed (e.g., file names and manifest contents), it mitigates risk by using robust, structured parsing (JSON and regex) rather than execution. Additionally, the agent is instructed to present the proposed plan to the user for confirmation before performing any write operations or test execution, maintaining human-in-the-loop oversight.

verifier-installer