efficiency
Skill0 starsChanged 41 days agoMITView on GitHub
One consolidated Claude/token efficiency dashboard — shells out to codeburn (spend, waste, model compare, yield), rtk (proxy savings), the model-router (routing mix), and cascade-brain (weak/dead skills), then synthesizes a single read. TRIGGER when user mentions "token efficiency", "where are my tokens going", "claude efficiency check", "token waste", "model usage", "routing mix", "efficiency dashboard", "am I wasting tokens", "claude code health". SKIP: tuning a single skill's description (use cascade-brain sharpen), compressing Claude's output (use caveman), discovering new tools to install (use cascade-discover).
1 other project uses a version of this.
Adopted by
Most recently taken up by
- zedarvates/frugal-skillstheir own version, first seen Nov 8, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- zedarvates/frugal-skills★ 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 repository42 other files it gives its agents.
- skills/composition-patterns/AGENTS.mdAGENTS.md
- skills/react-view-transitions/AGENTS.mdAGENTS.md
- skills/vercel-optimize/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- skills/last30days/CLAUDE.mdCLAUDE.md
- banner-designSkill
- brandSkill
- cascade-brainSkill
- cascade-discoverSkill
- cavemanSkill
- cdt-a11y-debuggingSkill
- cdt-chrome-devtools-cliSkill
- cdt-chrome-devtoolsSkill
- cdt-cookie-debuggingSkill
- cdt-debug-optimize-lcpSkill
- cdt-memory-leak-debuggingSkill
- cdt-troubleshootingSkill
- vercel-composition-patternsSkill
- core-web-vitalsSkill
- designSkill
- design-systemSkill
- email-best-practicesSkill
- frontend-slidesSkill
- grill-with-docsSkill
- humanizerSkill
- improve-codebase-architectureSkill
- job-search-kitSkill
- last30daysSkill
- react-emailSkill
- vercel-react-view-transitionsSkill
- seo-localSkill
- seo-mapsSkill
- slidesSkill
- ui-stylingSkill
- ui-ux-pro-maxSkill
- vercel-optimizeSkill
- web-design-guidelinesSkill
- web-quality-auditSkill
- wq-accessibilitySkill
- wq-best-practicesSkill
- wq-performanceSkill
- wq-seoSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 20
Every section is guarded — a missing tool prints ⚠ <tool> not installed — skipping and the rest still runs.
- Distinctive: line 24
Do not just paste the raw output.
- Distinctive: line 31
Keep the synthesis short and decision-oriented.
- Distinctive: line 57
Read-only end to end — the script only reads telemetry and runs reports.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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
Efficiency — one consolidated token/efficiency dashboard
A thin orchestrator. It does not re-implement any analytics — it shells out to the CLIs you already have (codeburn, rtk, the model-router, cascade-brain) via one script, then you synthesize the raw feed into a single read. The skill body stays deliberately thin: the script gathers, the model reasons.
Primary command
bash ~/.claude/skills/efficiency/dashboard.sh # FAST — 5 sections, safe to run anytime
bash ~/.claude/skills/efficiency/dashboard.sh --deep # FAST panel + 3 heavy codeburn analyses (slow)Fast panel (always): spend (codeburn 30d), proxy savings (rtk), model routing mix, skill-stack health (weak/dead descriptions), enforcement-layer presence checks (router / RTK / caveman).
Deep panel (--deep only): token-waste audit, per-model compare, shipped-vs-reverted yield. Skip --deep for a quick check — it runs the slow codeburn optimize/compare/yield passes.
Every section is guarded — a missing tool prints ⚠ <tool> not installed — skipping and the rest still runs.
After running: synthesize (this is the point)
Do not just paste the raw output. Run the script, then collapse it into a tight consolidated read with these four parts:
- Where tokens are going — spend trend (today vs month), rtk savings %, and which command classes dominate (e.g.
rtk readvs curls). Call out the biggest line items. - What's not firing — from the skill-stack audit, the dead/weak skills (no TRIGGER/SKIP, name-not-in-desc, very short). Surface the worst offenders by name; these are routing dead weight.
- Are the passive layers earning their keep — judge the model-router mix (is Haiku/Sonnet/Opus split sane, or is Opus over-firing?), the rtk savings rate, and whether router/RTK/caveman are all present (✓). Flag any layer that's installed but not paying off.
- What to cut / fix — a short prioritized list: descriptions to sharpen, routing thresholds to tune, waste to eliminate (cite the
--deepfindings if it was run).
Keep the synthesis short and decision-oriented. The user wants the forest, not every tree.
When to invoke
- "where are my tokens going" / "am I wasting tokens" / "token efficiency check"
- "claude code health" / "efficiency dashboard" / "model usage / routing mix"
- "token waste" → run with
--deepto fold in the codeburn optimize/compare/yield analyses - periodic check (e.g. weekly) on whether the passive efficiency layers are still pulling their weight
Skip when
- Sharpening ONE skill's description, or auditing descriptions to then edit them → use
cascade-brain(sharpen/audit); this skill only surfaces the weak list, it doesn't fix it. - Compressing Claude's own output to save tokens in-session → use
caveman. - Finding net-new tools to install → use
cascade-discover.
Files this skill touches
| Path | Role |
~/.claude/skills/efficiency/dashboard.sh | The gatherer — shells out to every CLI below; the only logic in this skill |
codeburn (CLI, on PATH) | Spend (status), and deep: optimize / compare / yield |
rtk (CLI, on PATH) | Proxy token savings (rtk gain) |
~/.claude/hooks/model-router.cjs | Routing mix (--stats); also presence-checked in the enforcement layer |
~/.claude/skills/cascade-brain/brain.py | Weak/dead skill audit (audit --weak-only) |
~/.claude/skills/caveman | Presence-checked in the enforcement layer |
Read-only end to end — the script only reads telemetry and runs reports. It never writes to skill files or state.
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.