agent-token-saver
Skill8 starsChanged 4 days agoMITView on GitHub
Reduce repeated work and noisy context while preserving task evidence. Use for long check output, exact log questions and RTK comparisons; skip already-small lookups.
In this repository3 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- agent-token-saverSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 13
Ask one exact question. Use scoped rg --no-config, jq or SQL before broad reads.
- Very specific: line 19
Use ats --json ... for scripts.
- Distinctive: line 20
Stop after acceptance. Write reference-only handoffs at real context boundaries.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Agent Token Saver
Optimize accepted work per total effort. Keep the task, permissions and oracle.
- Reuse decisions and green checks only with matching source evidence.
- Ask one exact question. Use scoped
rg --no-config,jqor SQL before broad reads. ats verify -- COMMANDexecutes once and preserves exit status plus full raw logs.ats project FILE --match TEXTfinds literal matches with line numbers and coverage counts.- Tail views and shortened inventories are previews. Never feed them to a parser expecting complete input.
ats bench --out DIRchecks eight synthetic RTK contracts. Reject evidence loss and zero byte gain.ats cost FILEuses disjoint usage and explicit prices. Bytes removed are not provider savings.- Use
ats --json ...for scripts;ats guidefor the next task. Load one primary skill per phase. - Stop after acceptance. Write reference-only handoffs at real context boundaries.
RTK is optional. Do not rewrite commands automatically, install hooks, spawn workers, make paid calls, start maintenance or change memory because this skill is loaded. Missing evidence stays unknown. The 20-skill catalog contains distilled principles, not an instruction to load every source.
The CLI, installer, fixtures and tests use TypeScript/Bun. Retired Python, shell, provider and hook surfaces are documented in docs/MIGRATION.md. A source update does not migrate old hooks already installed in a user's home.
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.

