The Agent Skills Directory

[SAFE]: The skill establishes a governance framework for token conservation, defining a 'read budget' and requiring the agent to stop and seek user approval before performing additional file reads.
[SAFE]: The instructions guide the agent to use efficient platform-native tools (Read, Grep, Glob) and built-in session commands (/status, /new, /compact) to maintain context hygiene.
[SAFE]: No suspicious activities such as remote code execution, sensitive data exfiltration, or prompt injection were detected.
[SAFE]: References to external capabilities, such as Qwen MCP delegation, are presented as optional optimizations for compute-intensive tasks without introducing unverifiable dependencies.

token-conservation