Agent is efficient and disciplined
phoenixlucky/zerotoken-skill · skills
Make Agent efficient and disciplined - not only save tokens: ZeroToken compresses invalid context/reasoning/output
★ 15GitHub stars
0Forks
2026-08-14Last updated
PythonLanguage
GPL-3.0License
Key features
- Based on your request characteristics, ZeroToken Skill automatically matches six task modes.
- All detailed specifications are defined in SKILL.md, and supporting script tools are in the scripts/ directory
- 1: Classify first, then budget: after receiving a request, first determine the mode and then allocate tokens
- Let the Agent do the most accurate thing with the fewest tokens
- Each mode has its own tool chain, output format and token budget strategy
Requirements
- Scenarios such as teaching/learning, brainstorming, in-depth research, detailed user requirements, etc. automatically exit ZeroToken mode and switch to detailed mode.
- See SKILL.md "When not to use ZeroToken" for complete rules.
Install command
dsh plugin --profile web add github:phoenixlucky/zerotoken-skill