More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Plan Usage Guard: Check Claude Code Limits Before Long Jobs

    1

    Don't lose a long Claude Code job to a usage limit halfway. Claude checks your Pro or Max limits first: GO, WAIT or STOP.

    Free

    1 useSecurity scanned
    Plan Usage Guard: Check Claude Code Limits Before Long Jobs

    Plan Usage Guard: Check Claude Code Limits Before Long Jobs

    Example session with this skill installed

    Before I kick off a big overnight refactor across about 40 files, check whether my plan limits can handle it.

    • Read your context and instructions
    • Compiled the plan usage guard
    Plan check: GO
    - Session (5-hour): 0% used, resets 6:20 PM CT (in 1h 54m)
    - Weekly: 1% used, resets Sun 5:00 AM CT
    

    The weekly limit has plenty of room. The session limit resets at 6:20 PM CT, so an overnight run will start on a fresh window from there. The reading is from 4:26 PM CT, so it's current.

    Session risk: A 40-file refactor with subagents can use up one 5-hour window. If it does, the run stops partway until the next reset.

    Auto-pause: It's off, so nothing will stop the run on its own.

    Plan: Split the refactor into batches that each leave the files finished. Re-check usage between batches. If a check shows the session near its limit, finish the current batch, write a progress note in the project, and wait for the reset.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    The problem. A long Claude Code job fails in the worst place when a plan limit runs out partway through: a refactor half applied, a subagent wave cut off, an overnight run that stopped at 2 AM. You find out in the morning and have to work out which files finished. With this skill, Claude checks your 5-hour session limit and your weekly limit before it starts, tells you plainly whether the job fits, and stops at a clean point with a resume note if a limit gets close mid-job.

    What it does. Ask "Before I kick off a big overnight refactor across about 40 files, check whether my plan limits can handle it." Claude answers in one block: GO, the session limit at 0% (resets 6:20 PM CT, in 1h 54m), the weekly limit at 1% (resets Sun 5:00 AM CT), and a plan to split the refactor into batches that each leave files finished, with a usage check between batches.

    • A verdict, not a percentage. GO, GO with checkpoints, WAIT or STOP, from a fixed table, so you don't have to judge whether 62% is enough for tonight.
    • The reset time for each limit, in your time zone, so a WAIT tells you exactly when to come back.
    • Re-checks between phases and before spawning subagents, since subagents spend from the same limits, so a job that started with room does not run dry unnoticed.
    • A clean stop. Near a limit it is built to finish the step in progress, write a progress note in your project, and hand you one line to send after the reset, so you resume instead of reconstructing.
    • Your thresholds. Defaults are 90% weekly and 85% session. Change them with one line in your CLAUDE.md.

    Why this beats prompting it yourself. The skill carries the rules you would otherwise restate every time: when to check, which limit matters for which job (a weekly limit takes days to reset, a session limit hours), what each level means, and how to stop without leaving a file half edited. Live figures come from the free, open-source plan-usage mod, which reads the limits Claude Code already receives with each reply: no network requests of its own, no model call. Without the mod, Claude asks you for the numbers from /usage and applies the same rules.

    Limits. Claude Code only, on a Pro or Max plan. Figures are as current as Claude Code's last reply, so use in a browser or on another machine shows up after the next reply. It checks before the job and between phases, not continuously; for a hard stop at a set percentage, turn on the mod's optional pause.

    What's inside: SKILL.md (when to check, the decision table, how to stop cleanly, the output format), references/tool-output.md (every field the usage tool returns), examples/scenarios.md (six worked cases, from plenty of room to the mod not installed).

    Free, and so is the mod it reads from.

    • Claude Code only, on a Pro or Max plan. Other agents and API keys have no plan limits to read.
    • Live figures need the free plan-usage mod. Without it, you read the numbers off /usage and tell Claude.
    • Figures are as current as Claude Code's last reply. Use in a browser or on another machine is not counted until the next reply.
    • It checks before a job and between phases, not after every step. The stop rules are instructions to Claude; for a hard stop at a set percentage, use the mod's optional pause.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    1 use

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    • Free with an Agensi account

    Permissions

    Write Files

    Instructions only: no scripts, no network access, no credentials or environment variables. The only write is a short progress note in your project when a long job pauses near a limit, so you can resume it.

    Listed1 month ago

    What's inside

    Frequently Asked Questions

    No. Session and weekly plan limits are a Claude subscription feature that only Claude Code reports.

    For automatic checks, yes (free, MIT). Without it, Claude asks you for the numbers from `/usage`.

    The mod answers locally with no model call. Like any tool, its name and description ride along with each request.

    It checks before the job and between phases, and when a check says WAIT or STOP it is built to finish the current step, write a progress note and stop there. For an automatic stop at a set percentage, turn on the mod's optional pause; the skill keeps the work in resumable steps around it.

    Yes. Add a line to your `CLAUDE.md`, such as `Plan usage guard: stop long work at 80% weekly and wait at 75% session.`

    New in LLM Ops