The Agent Skills Directory

[COMMAND_EXECUTION]: The skill manages an iterative optimization process by executing local commands. scripts/ar_run.sh executes the local benchmark script ./autoresearch.sh to measure performance metrics, while scripts/ar_log.py uses subprocess.run to call git for committing improvements and reverting regressions. Subprocess calls are implemented safely using argument lists, avoiding shell injection vulnerabilities.
[DATA_EXFILTRATION]: The skill does not perform any network operations. All performance data and session history are stored locally in autoresearch.jsonl and autoresearch.md.
[CREDENTIALS_UNSAFE]: No sensitive information, such as API keys or private tokens, is hardcoded or accessed by the skill scripts.

autoresearch