quota
Skill1 starsChanged 33 days agoMITView on GitHub
Read the model usage windows with `quota` and decide which model spends which window: GLM 5.3 first, then Codex, for per-change reviews; every Codex 5-hour window filled with the strongest model on queued review and audit tasks; Codex reset credits spent only once the weekly limit is gone. Use before launching a review, audit or delegated run, when choosing between GLM and Codex, when a Codex window or weekly limit is spent, when deciding whether to spend a reset credit, and to keep Codex windows busy with a queue of tasks (`quota-queue`).
2 other projects use a version of this.
Adopted by
Most recently taken up by
- sachingulati/claude-statuslinetheir own version, first seen Nov 1, 2026
- alexjmiller5/agent-config-publictheir own version, first seen Oct 2, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- sachingulati/claude-statusline★ 2same name
- alexjmiller5/agent-config-public★ 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 repository18 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- blast-radiusSkill
- claude-remote-control-serverSkill
- correctSkill
- councilSkill
- cross-reviewSkill
- docs-reviewSkill
- evidence-bundleSkill
- herdrSkill
- intent-cardSkill
- one-termSkill
- owner-charterSkill
- starbridgeSkill
- test-auditSkill
- unnarrateSkill
- unslopSkill
- verifySkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
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 9
Run quota before you pick a reviewer, launch an audit or queue work.
≈ “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
Quota
Run quota before you pick a reviewer, launch an audit or queue work. It
takes about 10 s and appends each reading to
~/.local/state/quota/history.jsonl.
codex 5h 100% 1:20 AM
codex week 43% Oct 3 at 7:20 PM 11% in deficit | Expected 32% used | Runs out in 2d 23h
codex reset credit available, expires 2026-10-04
zai 5h 21% 4:40 AM 5% in reserve | Expected 26% used | Lasts until reset
claude week 85% resets Sep 30, 10pm 12% in deficit | Expected 73% used | Runs out in 21h 43m
mistral api 93% €23.76 / €25.50 · €1.74 left
mistral vibe 37% €93.66 / €255.00 · €161.34 leftWhat the rows mean
- The percentage is how much of that window is used. The next column says when it resets, in the machine's local time.
- Pace compares use with an even spend across the window. "5% in reserve" means that much is unused so far and is lost at reset unless something spends it. "11% in deficit" means use runs ahead of even pace, and "Runs out in" says when the window hits 100% at the current rate. On the Codex weekly row a deficit is what Tom wants, because he wants the weekly limit used up.
- codex has a 5-hour window inside a weekly limit. Each
reset creditline is one credit and its expiry date. A credit is a full reset. It clears the weekly and the 5-hour window at once, and restarts the weekly timer: the next reset moves to 7 days after the spend. - zai is GLM 5.3 through opencode. It has a 5-hour window and no weekly cap, so headroom in one window is gone at the next.
- mistral api and mistral vibe are not headroom. Tom cancelled the Vibe plan; its last €8.50 of credit is not for reviews or any delegated run, so room on either row is never a reason to schedule work.
- claude is Tom's Claude subscription, which this session and its subagents spend. It is not a reviewer. A weekly row near 100% means fewer or cheaper subagents (Sonnet), not a different reviewer. A Claude reset refills the percentage but keeps the reset date, so the refill is lost at that date unless spent.
Who spends what
- Per-change reviews (
cross-review): GLM 5.3 through opencode, then Codex (gpt-6.1-sol) only when GLM is spent or failed. GLM goes first because its window has no weekly cap. Never Vibe. - A window about to reset with room left goes first. A zai window with reserve and under an hour to go, or a Codex weekly limit resetting within a day. Spend that one before the usual order.
- Fill every Codex 5-hour window with the strongest model
(gpt-6.1-sol at high effort; never gpt-6-astra, which spends a whole 5-hour window before one run finishes) on deep reviews and audits, through
quota-queue. Room left when a window closes is lost. This is separate from step 1. Routine reviews don't wait for Codex, and deep reviews don't wait for GLM. - Reset credits only once the weekly limit is at 100%, the credit that expires first. At 5-hour 100% with weekly room, wait for the 5-hour reset, because a credit spent then clears a weekly limit that still had room. At weekly 100% with the natural weekly reset under a day away, wait for it unless the credit expires first: the spend pushes the next reset 7 days out, so it buys only the hours until the natural one.
Helpers on PATH
codex-limitsprints<5h-used-%> <weekly-used-%> <5h-resets-at-epoch> <credits>in under a second, e.g.100 43 1790637613 3. Poll this, notquota.codex-limits --spendspends the credit that expires first and printsreset,nothingToReset,noCreditoralreadyRedeemed.quota-queue <dir>runs the executables in<dir>/pending/in name order while Codex has room, sleeps through a spent window, and spends a credit at weekly 100%. Run it withrun_in_background; it exits withQUEUE EMPTY, which is your cue to add tasks. Task layout, what to queue and the log: reference/queue.md.
Gotchas
- You tested
--spend. A spend cannot be undone and credits are scarce. Check with plaincodex-limits; spend only under rule 4. - You looked for a Codex CLI command that spends a credit. There is
none. Only
codex app-server(account/rateLimitResetCredit/consume) does, which is whycodex-limitsexists. - A queued task never ran.
quota-queueruns only files with the executable bit;chmod +xeach task. another quota-queue runs on <dir>. One runner per directory holds a lock. Don't start a second; add tasks topending/and the running one picks them up.- A task ran against a stale PR head. Build the worktree inside the task, not when you queue it, because it may run hours later.
codex-limitsprintedNone. The account is logged out or the app-server changed its reply.quota-queuereads that as unreadable and retries every 5 minutes; post Tom a card instead of spending.- A credit will expire before the weekly limit can run out. Tom decides: post a card with the expiry date. Don't spend it early.
Not possible
- Resetting only the 5-hour window. Every credit clears both.
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.

