More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    how-long: Live Progress Bar & ETA

    1

    Claude publishes a live progress page for long tasks: steps, a segmented progress bar, time elapsed and an ETA that learns from real timings.

    $1

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    how-long: Live Progress Bar & ETA

    how-long: Live Progress Bar & ETA

    Example session with this skill installed

    Go through the 12 CSV files in ./sales-2025: clean them, compute the quarterly metrics, draw the charts and write up the Q3 report. It'll take a while, so show me a progress bar and how long it will take.

    • Read your context and instructions
    • Compiled the how-long
    • Generated the document

    Summary

    Live progress is on. I published a page called

    "Q3 report progress": open its card in this conversation at any time to see what I'm doing and how long is left. I update it at every step.

    The progress page, 25 minutes in

    how-long progress page

    #StepStatusTimeEstimate
    1Read and check 12 data filesDone (2 files were GBK-encoded, converted)4 min3 min
    2Clean the dataRunning, 7 / 12 files21 min8 min
    3Compute quarterly metricsPending5 min
    4Draw chartsPending6 min
    5Write the reportPending10 min
    6Export PDFSkipped: switched to a Docs document2 min

    Time left: 56 min ·

    Expected to finish: 17:01 ·

    Progress: 24% ·

    Pace: 2.2× the estimated time ·

    Estimate quality: Fair

    Activity

    • Task started
    • 2 files had the wrong encoding, converted to UTF-8 (warning)
    • Step 1 done, cleaning started

    When it finishes

    The page switches to

    Done, shows the total time and a one-line result, for example: "Report done: 18 pages with 6 charts. Q3 revenue up 12% on Q2." If your agent supports push notifications, you get one as well.

    Next steps

    • Keep working on something else: the page updates by itself
    • Ask "how long is left?" at any time, or just open the card
    • Say "no progress page" to turn it off for the rest of the conversation

    screenshot-light-en.png

    PDF · document

    Generated

    Example file from a real run - the skill writes it into your workspace.

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

    About this skill

    Long agent tasks are a black box: you can't tell how far along Claude is or how much longer it will take. how-long fixes that. At the start of a long task, Claude publishes a live progress page you can open at any time.

    What you see

    • What Claude is doing right now
    • Every step, with its status and time taken
    • A segmented progress bar, each segment sized by that step's estimated time
    • Time elapsed, time left and the expected finish time
    • Whether the work is running faster or slower than estimated ("20% slower than estimated")
    • A short activity log

    The page updates live, with no refresh. It follows your language (English or Chinese) and your light or dark theme.

    When it kicks in

    • Tasks Claude expects to take about 10 minutes or more, or with 6 or more steps
    • Whenever you ask for a progress bar, progress updates, an ETA or "how long"
    • Say "no progress page" and it stays off for the rest of the conversation

    How the ETA works

    Claude estimates each step when it plans the task. The page learns a pace from real timings (actual time compared with the estimates, smoothed so one early outlier doesn't swing it), uses the measured rate for steps like "7 / 12 files", and leaves time spent waiting for your reply out of the estimate. If updates stop for a long time, the page says the task may have stopped instead of claiming it is still running.

    Built for long runs

    Claude saves the page address in progress-state.json, so if its context is compacted halfway through, it finds the page again and keeps updating it.

    Requirements

    The live page uses Claude's artifacts with a database (the Artifact and ArtifactData tools): the Claude app and Claude Code sessions that have artifacts. In agents without them, the skill falls back to the agent's own task list or a checklist, so you still get step tracking, but no live page or ETA.

    Open source (MIT): https://github.com/bhu0345/claude-how-long Live demo: https://bhu0345.github.io/claude-how-long/en/

    The live page needs Claude's Artifact and ArtifactData tools; in other agents the skill falls back to the task list. Each checkpoint costs Claude one or two extra tool calls, so it is meant for tasks of about 10 minutes or more, or when you ask for it. It uses a shell with python3 once per task to copy the page template. The page is in English or Chinese, following the language you write in.

    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

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    Requires

    Claude Artifacts
    Terminal access
    Python 3

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access

    Terminal / Shell: runs `date +%s` for timestamps and a one-line python3 command that copies template.html with the task title filled in. Read files: reads the skill's own template.html. Write files: writes the progress page (progress-<task>.html) and progress-state.json in the working directory. Network: the progress page loads IBM Plex fonts from Google Fonts when you open it. The skill itself makes no network requests.

    Listed1 month ago

    What's inside

    Frequently Asked Questions

    The live page needs Claude's artifacts with a database (the Artifact and ArtifactData tools), which the Claude app and Claude Code sessions with artifacts have. Other agents don't have them, so the skill falls back to their task list or a checklist: you get step tracking, but no live page or ETA.

    For tasks Claude expects to take about 10 minutes or more, or with 6 or more steps, and whenever you ask for a progress bar, progress updates or an ETA. Say "no progress page" and it stays off for the rest of the conversation.

    Claude estimates each step up front. The page then learns a pace from real timings, smoothed so one step can't swing it, and uses the measured per-item rate when a step has counts like 7 / 12 files. Until the first step finishes the estimate is labelled Rough; it moves to Fair and Good as real timings come in.

    When Claude publishes the page it saves its address in progress-state.json. After compaction it reads that file and the page's stored state, then keeps updating the same page.

    The page notices. With no update for 1.5 times the running step's estimate (at least 6 minutes) it shows a warning, and at three times that it says the task may have stopped, instead of showing "running" forever.

    Two files and no dependencies: SKILL.md (the instructions Claude follows, about 9 KB) and template.html (the progress page, copied once per task). The page shows only what Claude writes to it, such as the task title and step names, and it is private to you unless you share it.

    New in Workflow Automation