The Agent Skills Directory

[SAFE]: The skill is designed to improve agent performance by identifying context bloat. It deterministicly scans local files like settings.json, CLAUDE.md, and skills to calculate token usage.
[COMMAND_EXECUTION]: The skill utilizes two Python scripts, scripts/audit.py and scripts/baseline.py, to perform its analysis. These scripts are implemented using the Python standard library, requiring no external packages, and do not perform network requests or unauthorized system modifications.
[SAFE]: Data access is confined to the agent's own configuration and instruction files within the project and user home directories. The skill does not attempt to access credentials, SSH keys, or other sensitive system data unrelated to its auditing purpose.
[SAFE]: The skill follows security best practices by presenting all proposed configuration changes to the user as diffs for review and manual approval, rather than auto-applying them.

context-audit