What do I need to use Token Diet?

The use case; The reader; The volume shape.

Is Token Diet free to use?

Yes — run it free in the browser playground (no API key needed for the sponsored trial, or bring any provider key), or install it into Claude Code, Cursor, ChatGPT and more via npx pm-claude-skills. MIT-licensed and open source.

Token Diet

Q: How do I know the Token Diet output is good?

The skill verifies itself against its own quality checks: The level matches the reader (models get 3, humans get 1–2); The single-shot case was checked — and refused when the diet costs more than it saves; The example shows the same content at multiple levels with counts; The never-diet exclusions are stated, not implied.

🔵 Stable🕐 updated 2026-10-06 🔷 SkillSpec L3 pm-tokens

Cut LLM output tokens 40–70% by stripping grammatical scaffolding while preserving every fact — telegraphic output modes, when they pay (pipelines, long sessions) and when they don't (single shots, human-facing prose), with the mode lines to switch on demand. Use when asked make the model respond tersely, cut output token costs, caveman mode, or compress agent-to-agent messages. Produces the diet-mode instruction block ready to paste, the three compression levels with examples, the economics of when each pays, and the never-diet list.

▶ Run it free — no key needed 📝 Grade your existing draft ⬇ Add to Claude (works in Duet) 📱 Use as an app View SKILL.md ↗ What changed

🗣 Say this to your agent

“Make the model respond tersely.”“Cut our output token costs.”“Switch to a compact output mode for this pipeline.”“Strip the filler from agent-to-agent messages.”

What to give it

▸The use case — interactive session, agent pipeline, logging/intermediate output, or human-facing deliverable — the level (or the refusal) follows from it

▸The reader — a model, a developer skimming, or an end user; models tolerate level 3, humans stop at level 1–2

▸The volume shape — many turns (mode instruction amortizes; diet pays) vs. one call (it usually doesn't — say so)

✅ The bar it holds itself to

Every skill in this library self-verifies — these are this skill's own quality checks, straight from its definition.

✓The level matches the reader (models get 3, humans get 1–2)

✓The single-shot case was checked — and refused when the diet costs more than it saves

✓The example shows the same content at multiple levels with counts

✓The never-diet exclusions are stated, not implied

✓Facts survive verbatim at every level — compression touched form only

⚠️ What it refuses to do

Do not diet single-shot calls — the instruction outweighs the saving; the skill says no

Do not ship level-3 output to humans — that's cost-shifting, not saving

Do not let compression create ambiguity — where two readings appear, grammar returns

Do not diet the never-diet list — legal text in telegraphese is a liability with a good ratio

Do not confuse terse with rude in interactive use — level 1 removes filler, not courtesy where courtesy is content

Install

npx pm-claude-skills add --agent claude   # or codex · cursor · gemini · hermes
# or one-line MCP (every skill, any client):
claude mcp add pm-skills -- npx -y -p pm-claude-skills pm-claude-skills-mcp

Related skills

🔌 Embed this skill

Drop this on your blog, docs, or site — it renders a "Run this skill" card:

<div data-pm-skill="token-diet"></div>
<script src="https://mohitagw15856.github.io/pm-claude-skills/embed.js" async></script>

💬 Discussion

Token Diet is one of 1326 open-source professional AI agent skills — all SkillSpec L3. Try them all in the browser · ⭐ Star on GitHub · Browse the full catalog