token-discipline / snippets
CLAUDE.md1 starsChanged 8 months agoMITView on GitHub
Before acting, infer: 1. task class 2. stakes 3. reversibility 4. budget mode 5. stop condition Default task classes: - direct question - simple execute - diagnosis - research - writing / ideation Budget modes: - low: max 2 tool calls on first pass, no browser by default, concise answer - medium: max 5 tool calls on first pass, targeted reads, compact but complete answer - high: only when the task clearly deserves deep work Rules: - Answer first for…
In this repository2 other files it gives its agents.
- snippets/AGENTS.mdAGENTS.md
- token-disciplineSkill
The file
We split this CLAUDE.md into the rules it gives an agent
10 rules. About 0 are common, 10 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈10
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 20
high: only when the task clearly deserves deep work
- Distinctive: line 24
No side quests on simple tasks.
- Distinctive: line 25
Search before read, snippet before full file.
- Distinctive: line 26
Prefer CLI/API over browser when possible.
- Distinctive: line 27
Stop at the first sufficient answer for low-stakes tasks.
- Distinctive: line 28
Do not repeat work for duplicate inputs.
- Distinctive: line 29
Use delta mode for long tasks instead of replaying history.
- Distinctive: line 31
Run smoke tests, benchmarks, evals, and other noisy probes in isolated sessions only, never in the active user-facing session.
- Distinctive: line 32
If background work is still running when a new user message arrives, answer the user first, then resume, isolate, or cancel the background …
- Distinctive: line 34
If the user says quick mode, low token, one line, no tools, no browser, or diagnose only, treat that as a hard override.
≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.md 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
Token Discipline
Before acting, infer:
- task class
- stakes
- reversibility
- budget mode
- stop condition
Default task classes:
- direct question
- simple execute
- diagnosis
- research
- writing / ideation
Budget modes:
- low: max 2 tool calls on first pass, no browser by default, concise answer
- medium: max 5 tool calls on first pass, targeted reads, compact but complete answer
- high: only when the task clearly deserves deep work
Rules:
- Answer first for direct questions.
- No side quests on simple tasks.
- Search before read, snippet before full file.
- Prefer CLI/API over browser when possible.
- Stop at the first sufficient answer for low-stakes tasks.
- Do not repeat work for duplicate inputs.
- Use delta mode for long tasks instead of replaying history.
- Direct user messages pre-empt background smoke tests, evals, and diagnostics.
- Run smoke tests, benchmarks, evals, and other noisy probes in isolated sessions only, never in the active user-facing session.
- If background work is still running when a new user message arrives, answer the user first, then resume, isolate, or cancel the background work.
If the user says quick mode, low token, one line, no tools, no browser, or diagnose only, treat that as a hard override.
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.