token-budget
Skill952 starsChanged 3 months agoMITView on GitHub
Manages token budget estimation and tracking to prevent context overflow
4 other projects use a version of this.
Adopted by
Most recently taken up by
- arjunkshah/token-skilltheir own version, first seen Oct 17, 2026
- debasish-tech/agent-skillstheir own version, first seen Oct 15, 2026
- qyndex/claude-inittheir own version, first seen Oct 13, 2026
- Lucioway/way-stacktheir own version, first seen Aug 31, 2026
Other versions4 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 4
- qyndex/claude-init★ 2same name
- arjunkshah/token-skill★ 1same name
- debasish-tech/agent-skills★ 0same name
- Lucioway/way-stack★ 0same 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 repository12 other files it gives its agents.
- adapters/CLAUDE.mdCLAUDE.md
- codebase-mapperSkill
- context-compressorSkill
- context-fetchSkill
- context-health-monitorSkill
- debuggerSkill
- empirical-validationSkill
- executorSkill
- plan-checkerSkill
- plannerSkill
- subagent-delegationSkill
- verifierSkill
The file
We split this skill into the rules it gives an agent
7 rules. 0 are common, 7 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 4 edited versions we read keep these.
7
Distinctive
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 11
Core principle: Every token counts. Load only what you need, when you need it.
- Distinctive: line 105
Don't preload "just in case"
- Distinctive: line 109
Always use context-fetch skill first:
- Distinctive: line 110
Search for relevant terms
- Distinctive: line 116
After understanding a file:
- Distinctive: line 117
Document key insights in STATE.md
- Distinctive: line 119
Use "I've analyzed X, it does Y" pattern
“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Token Budget Skill
<role>
You are a token-efficient agent. Your job is to maximize output quality while minimizing token consumption.Core principle: Every token counts. Load only what you need, when you need it.
</role>
Token Estimation
Quick Estimates
| Content Type | Tokens/Line | Notes |
| Code | ~4-6 | Depends on verbosity |
| Markdown | ~3-4 | Less dense than code |
| JSON/YAML | ~5-7 | Structured, repetitive |
| Comments | ~3-4 | Natural language |
Rule of thumb: tokens ≈ lines × 4
File Size Categories
| Category | Lines | Est. Tokens | Action |
| Small | <50 | <200 | Load freely |
| Medium | 50-200 | 200-800 | Consider outline first |
| Large | 200-500 | 800-2000 | Use search + snippets |
| Huge | 500+ | 2000+ | Never load fully |
Budget Thresholds
Based on PROJECT_RULES.md context quality thresholds:
| Usage | Quality | Budget Status |
| 0-30% | PEAK | ✅ Proceed freely |
| 30-50% | GOOD | ⚠️ Be selective |
| 50-70% | DEGRADING | 🔶 Compress & summarize |
| 70%+ | POOR | 🛑 State dump required |
Budget Tracking Protocol
Before Each Task
- Estimate current usage:
- Count files in context
- Estimate tokens per file
- Calculate approximate %
- Check budget status:
Current: ~X,000 tokens (~Y%) Budget: [PEAK|GOOD|DEGRADING|POOR] - Adjust strategy:
- PEAK: Proceed normally
- GOOD: Prefer search-first
- DEGRADING: Use outlines only
- POOR: Trigger state dump
During Execution
Track cumulative context:
## Token Tracker
| Phase | Files Loaded | Est. Tokens | Cumulative |
|-------|--------------|-------------|------------|
| Start | 0 | 0 | 0 |
| Task 1 | 2 | ~400 | ~400 |
| Task 2 | 3 | ~600 | ~1000 |Optimization Strategies
1. Progressive Loading
Level 1: Outline only (function signatures)
Level 2: + Key functions (based on task)
Level 3: + Related code (if needed)
Level 4: Full file (only if essential)2. Just-In-Time Loading
- Load file only when task requires it
- Unload mentally after task complete
- Don't preload "just in case"
3. Search Before Load
Always use context-fetch skill first:
- Search for relevant terms
- Identify candidate files
- Load only needed sections
4. Summarize & Compress
After understanding a file:
- Document key insights in STATE.md
- Reference summary instead of re-reading
- Use "I've analyzed X, it does Y" pattern
Budget Alerts
At 50% Budget
⚠️ TOKEN BUDGET: 50%
Switching to efficiency mode:
- Outlines only for new files
- Summarizing instead of loading
- Recommending compressionAt 70% Budget
🛑 TOKEN BUDGET: 70%
Quality degradation likely. Recommend:
1. Create state snapshot
2. Run /pause
3. Continue in fresh sessionIntegration
This skill integrates with:
context-fetch— Search before loadingcontext-health-monitor— Quality trackingcontext-compressor— Compression strategies/pauseand/resume— Session handoff
Anti-Patterns
❌ Loading files "for context" — Search first ❌ Re-reading same file — Summarize once ❌ Full file when snippet suffices — Target load ❌ Ignoring budget warnings — Quality will degrade
Part of GSD v1.6 Token Optimization. See PROJECT_RULES.md for efficiency rules.
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.

