The Agent Skills Directory

[SAFE]: The skill provides instructional content aimed at improving the reliability and safety of coding agents. It emphasizes surgical edits, defining success criteria before acting, and rigorous verification of results.
[SAFE]: The included validation script (scripts/validate-fixtures.js) performs local file reading and string matching to ensure the skill content remains consistent with its requirements. It does not perform network operations, use unsafe dynamic execution, or access sensitive files.
[SAFE]: External references to GitHub and X (Twitter) are for documentation purposes and do not trigger remote code execution or data exfiltration. The skill specifically highlights its ability to function offline by embedding necessary summaries.

karpathy-guidelines