Is the Token Discipline skill free?

Yes. The skill itself is free from Skillproofdev/token-discipline. SkillProof publishes the install command and an independent test verdict at no cost.

What is the Token Discipline SkillProof Score?

8.0/10 — installs cleanly 5/5, triggers reliably 4/5, output vs. baseline 6/10, docs & honesty 5/5.

How do I install Token Discipline?

Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.

Can I use Token Discipline with Cursor, Copilot, Gemini CLI, Codex or other AI tools?

The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.

Token Discipline

Input-side token rules for Claude Code — ~20% off multi-step work, wash on one-shots.

by SkillProof · Skillproofdev/token-discipline

Tested · Works

Install — copy the command Download source ↗

Test report

Verdict: Tested · Works
Score: 8.0/10
Tested: Jul 11, 2026
Environment: Claude Code 2.x (agent harness)
Upstream re-checked: Aug 10, 2026 · d58bb5a

Our own skill — tested under the same protocol, disclosed. README's raw telemetry (5 task pairs, baseline vs with-skill agent, identical prompts) shows a real but narrow win: −19.0% and −19.9% tokens on the two multi-step tasks (codebase Q&A, multi-file edit), versus −2.1%, +0.0%, +5.7% on the three one-shot tasks — a wash there, as the README itself admits. No independent bench/ directory exists; this verdict is sourced entirely from the README's own numbers, which is a limitation on top of the modest effect size. Protocol test caught real issues in our own repos (dead benchmark links, strict-YAML frontmatter) — fixed and re-verified same day.

Scored on four weighted criteria — install, triggering, output vs. baseline, docs. How scoring works

Installs cleanly 5/5
Triggers reliably 4/5
Output vs. baseline 6/10
Docs & honesty 5/5

What Token Discipline does

Enforces input-side token discipline in Claude Code sessions — scoped search before reads, sliced reads instead of whole files, batched independent tool calls, and no redundant re-reads after successful edits — plus terse final output. Benchmarked on 5 real task pairs against a baseline agent: cuts tokens ~19-20% on multi-step codebase work (editing, multi-question Q&A) but is essentially a wash on trivial one-shot tasks, a limitation the README discloses rather than hides.

Token Discipline before / after — same task, with and without

Without the skill

Baseline agent (no skill) on the 4-file coordinated-edit task read whole files to locate functions and re-read every edited file afterward to 'verify' the change — 68,219 tokens across 16 tool calls, with the transcript literally stating it "verified by re-reading the edited regions."

With Token Discipline

With token-discipline loaded, the same task used grep-then-slice reads and trusted the Edit tool's own error signaling instead of re-reading — 54,637 tokens (−19.9%) across 19 tool calls, producing an equivalent, correct diff (the two outputs were diffed to confirm no quality loss).

How to install Token Discipline

git clone https://github.com/Skillproofdev/token-discipline ~/.claude/skills/token-discipline

Skills live in ~/.claude/skills/ (global) or .claude/skills/ (per-project). Restart Claude Code after installing.

Commands — how to trigger Token Discipline

/token-discipline Input-side token rules for Claude Code — ~20% off multi-step work, wash on one-shots.

It also activates on plain-language prompts like these:

This session is burning tokens way too fast — can you tighten up how you're working?
We keep hitting the context limit on this long refactor. Work more token-efficiently for the rest of it.
Can you reduce Claude Code's token usage while we do this multi-file coordinated edit?

Frequently asked questions

Is the Token Discipline skill free?: Yes. The skill itself is free from Skillproofdev/token-discipline. SkillProof publishes the install command and an independent test verdict at no cost.
Does Token Discipline work with Claude Code?: We tested it with Claude Code 2.x (agent harness) on Jul 11, 2026. Verdict: Tested · Works. Our own skill — tested under the same protocol, disclosed. README's raw telemetry (5 task pairs, baseline vs with-skill agent, identical prompts) shows a real but narrow win: −19.0% and −19.9% tokens on the two multi-step tasks (codebase Q&A, multi-file edit), versus −2.1%, +0.0%, +5.7% on the three one-shot tasks — a wash there, as the README itself admits. No independent bench/ directory exists; this verdict is sourced entirely from the README's own numbers, which is a limitation on top of the modest effect size. Protocol test caught real issues in our own repos (dead benchmark links, strict-YAML frontmatter) — fixed and re-verified same day.
What is the Token Discipline SkillProof Score?: 8.0/10 — installs cleanly 5/5, triggers reliably 4/5, output vs. baseline 6/10, docs & honesty 5/5.
How do I install Token Discipline?: Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.
Can I use Token Discipline with Cursor, Copilot, Gemini CLI, Codex or other AI tools?: The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.