The Agent Skills Directory

[COMMAND_EXECUTION]: The skill heavily promotes the use of bash commands (e.g., sed, awk, grep, cp, mv, tail, wc) to perform file and data operations.
It specifically instructs the agent to use CLI tools instead of the native Read, Edit, and Write tools whenever possible to minimize token consumption.
Examples provided include using sed -i for in-place file editing, cat > file << EOF (heredocs) for file creation, and cp for file duplication.
These techniques are framed as 'Token Optimization Rules' and include guidance on handling shell aliases and platform-specific command differences (e.g., macOS vs. Linux sed).
[SAFE]: The analysis did not find any evidence of malicious patterns. The instructions focus on cost-effective model selection (Opus vs. Sonnet), efficient log handling, and project organization using the MANIFEST system. No prompt injection attempts to bypass safety filters or data exfiltration vectors were identified.

token-efficiency