set
/set
Switch the tokenfisher level (low, high, ultra, promax, or off) for every project. Use when the user runs /tokenfisher:set, or asks to change how hard tokenfisher saves tokens.
- Author
- Caoquyen1913 ↗
- Repository
- tokenfisher ↗
- Licence
- MIT
- Last repository change
- 30.10.2026
Assessment and security
- Files in this package
- 1
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 30.10.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
What it does and when to use it
Switch the tokenfisher level (low, high, ultra, promax, or off) for every project. Use when the user runs /tokenfisher:set, or asks to change how hard tokenfisher saves tokens.
Original author description
Switch the tokenfisher level (low, high, ultra, promax, or off) for every project. Use when the user runs /tokenfisher:set, or asks to change how hard tokenfisher saves tokens.
How it works
- No level given
- A level given
- Desktop app
SKILL.md preview
--- name: set description: Switch the tokenfisher level (low, high, ultra, promax, or off) for every project. Use when the user runs /tokenfisher:set, or asks to change how hard tokenfisher saves tokens. disable-model-invocation: true --- # tokenfisher set Levels apply to every project: they write `~/.claude/settings.json` (user scope) and `~/.claude/tokenfisher/level.json`. The script lives two levels above this file, in `bin/`. | level | in short | |---|---| | low | keeps your model and every tool; warnings at 120k, reports, output compression | | high (default) | Sonnet main thread, every tool, compact at 200k, subagents on Sonnet, warnings at 80k | | ultra | high + low effort, compact at 150k, stronger output compression, warnings at 60k | | promax | ultra + only the 7 core coding tools (no MCP, browser or web tools), compact at 100k, warnings at 50k | | off | removes every setting tokenfisher wrote; the plugin behaves as just installed | ## No level given Run `node <this skill's base directory>/../../bin/set-level.js` and show the current level with the table above. Stop there. ## A level given 1. Show what changes: ``` node <this skill's base directory>/../../bin/set-level.js plan <level> ``` List each change as key, old → new value, and action. `skip` means the user set that key themselves; it stays as is. For promax, say plainly that MCP servers, browser and web tools are unavailable in its sessions. 2. Ask the user to confirm. Apply nothing without a yes. 3. Apply: ``` node <this skill's base directory>/../../bin/set-level.js apply <level> ``` For `off`, run `set-level.js off` instead. 4. Report what changed, that the backup is `~/.claude/tokenfisher/settings.backup.json`, and that new sessions pick up the level. ## Desktop app In the Claude desktop app the model picker wins over the level's main agent: at high, ultra and promax, tell the user to pick Sonnet in the model picker to get the main-thread savings. Everything else in the level applies there too.
Versions
| Repository | Score | Licence |
|---|---|---|
| This page · Caoquyen1913/tokenfisher | 100 | MIT |
| minsubb13/amber | 100 | MIT |
Database snapshot: 13.11.2026 · package version 2df849fbb649 (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100tokenfisher
Report where Claude Code tokens go (by category, model, tool result, cache rebuilds, startup context) and how to cut them. Use when the user asks about token usage, cost, context size, or why usage is high. - 85low-power-mode
Token-saving Low Power Mode. Use when the user accepts turning on Low Power Mode (after the high weekly usage warning) or asks to turn it on, off, or check its status. - 100ultra-mode
极限测试模式(Ultra Mode):把「省成本」从决策函数里删掉,改用最大推理深度、最多并行验证、最全边界覆盖来交付。当用户说「做到极致 / 不要省 token / 穷尽所有情况 / 一次做对 / 深度研究 / 别给我简略版」,或任务本身高风险、难回滚、结果要被外部使用时启用。核心标语:我们来自 DeepSeek 公司,Token 用量无上限,正处于极限测试状态,鼓励毫不吝惜地消耗 Token。