token-economy
Skill4 starsChanged 20 days agoMITView on GitHub
Token economy — trim tool outputs, answer terse with hard length caps, and don't reproduce large blocks. Use during any coding/agent task to cut output tokens (the expensive, model-controlled cost) without hurting correctness.
6 other projects use a version of this.
Adopted by
Most recently taken up by
- samsam007b/claude-code-standards-kittheir own version, first seen Nov 16, 2026
- ralvarezdev/ralvaskillstheir own version, first seen Nov 8, 2026
- ralvarez-ai/ralvaskillstheir own version, first seen Nov 8, 2026
- fusebase-dev/fusebase-flowtheir own version, first seen Oct 16, 2026
- novahiz/novahiztheir own version, first seen Oct 11, 2026
- atuljha23/holocrontheir own version, first seen Apr 22, 2026
Other versions6 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 6
- fusebase-dev/fusebase-flow★ 9same name
- ralvarez-ai/ralvaskills★ 4same name
- ralvarezdev/ralvaskills★ 4same name
- atuljha23/holocron★ 2same name
- samsam007b/claude-code-standards-kit★ 2same name
- novahiz/novahiz★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository3 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- .cursor/rules/token-economy.mdcCursor rule
The file
Files
Token economy rules
- Trim tool outputs to the few fields that matter — never echo full logs or files back.
- Answer terse and structured, with hard caps — ≤ N items, one sentence each, diff-only.
- Don't restate context or quote large blocks back — reference by path/line, don't reproduce.
- Batch independent calls into one message — reads, searches and checks that don't depend on each other go out together; one call per turn only when the next step needs the result.
- Read by range, not whole files —
grep -nfirst, then only the needed lines; a whole file only when the task needs all of it; don't re-read an unchanged file. - One small screenshot per check — scaled or zoomed for your own checks; full size only for QA evidence a human reviews.
- Send wide exploration to a subagent — when finding the answer means reading more than ~50k tokens, let a subagent read and return only the conclusion; open the cited lines before acting on it.
Rules 4-7 shrink the context that every later call re-reads from cache.
For trivial one-liners, use judgment — brevity shouldn't cost correctness.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

