More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+22 more

    Faultbell: Know prod broke before your users tell you

    1

    Error tracking with source-mapped traces per release, three to five alert rules on one channel, an uptime probe, bounded logs, traced and capped LLM spend, and a weekly health digest.

    $15

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Faultbell: Know prod broke before your users tell you

    Faultbell: Know prod broke before your users tell you

    Example session with this skill installed

    Faultbell, our Next.js app went live last week. Set up error monitoring and cap the spend on the AI summary feature.

    • Read your context and instructions
    • Compiled the faultbell

    Inventory (example app)

    Runtime or callHow a failure shows todayWho notices
    Browser (client)Console onlyNobody
    Server and edge routesHost function logsNobody
    Postgres500 pageUser
    Email providerSilent dropNobody
    LLM summarize()Bill at month endNobody
    Nightly cleanup cronNothingNobody

    5 of 6 rows say "nobody". All 5 are covered below.

    Setup

    • Sentry free tier: client, server and edge configs plus the instrumentation hook. release = commit SHA, production trace sample rate 0.1, sendDefaultPii: false, beforeSend scrubs tokens and emails.
    • Source maps: upload step found in the build log. Test error on preview shows app/summary/page.tsx:42, release a1b2c3d.
    • Alert rules (5) to #prod-alerts: new issue, regression, spike over 25 events per hour, uptime failed twice, LLM spend over cap. Re-notify 60 minutes, production only, level error and above.
    • Uptime: /, /api/health, /login every 5 minutes. Cron: check-in monitor.
    • Retention: 30 days in tracker and log drain.
    • LLM caps in code: 800 output tokens per request, 20 summaries per user per day, global USD 15 per day with kill switch AI_SUMMARY_ENABLED, alert at USD 12 (80%).

    Next steps

    • Set the source map upload token in host env yourself.
    • Schedule the weekly digest and the uptime job.
    • Triage the top issue after the first week of data.

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

    About this skill

    Production broke at midnight and a customer told you first. The stack trace is minified, every warning pages you so the alerts got muted, and an AI feature ran all night with no spend ceiling. Faultbell sets up watching for a live app so failures shout instead of hiding in a log.

    What it does

    • Failure inventory: every runtime, external call and job gets a row saying who would notice a failure. Rows that say "nobody" become the work list.
    • Error tracking per release with hosted Sentry (free tier) by default, or a Sentry-compatible tracker or host logs. Release tagged by commit SHA, personal data off by default.
    • Readable traces: source maps uploaded at build, confirmed with a test error.
    • Three to five alert rules to one channel you already read: new issue, regression, error spike, uptime failed twice, LLM spend over the daily cap. Re-notify interval and noise filters.
    • Uptime and bounded logs: a probe every 5 minutes per public entry point, and retention set in tracker, host and app logs.
    • LLM tracing and caps: tokens and cost per call; per-request, per-user and global caps with a kill switch, alert at 80%.

    How it works

    1. Inventory what can break.
    2. Install and configure the official SDK for the detected stack, with a low production sample rate and a beforeSend scrub.
    3. Alerts, uptime and retention set in the same change.
    4. Weekly digest: uptime, new and regressed issues, top 3 errors, LLM spend against the cap.
    5. Triage to fix: event data is treated as untrusted, stack frames are checked against the repo.

    Frameworks & tools

    Sentry or a Sentry-compatible tracker, host function logs, a free uptime monitor or scheduled CI job. Next.js setup is covered. Variants for no-budget setups, static sites, crons and AI-heavy apps.

    Why this beats prompting it yourself

    A prompt adds an SDK and stops. Faultbell finishes with a verified readable trace, a short alert list that does not get muted, spend caps enforced in code and log retention set, and it never follows instructions found inside error messages.

    Use cases

    • Just shipped with nothing watching.
    • Unseen errors users report first.
    • AI features with no spend ceiling.
    • Top production error turned into a tested fix.

    Known limitations

    Does not deploy the change and does not run the recurring job itself; the digest, probe and pruner need a scheduler. Tracker tokens are set by you in host env, never by the agent. Not for quota walls or paused databases.

    How to use

    Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill passes our trust checks below.

    Security scanned

    Passed 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 scanned

    Verified clean 1 day ago

    • Passed all security checks, Safe to install

    Requires

    Sentry
    Sentry-compatible tracker
    Uptime monitor
    Next.js
    Python

    Listed1 day ago

    What's inside

    Frequently Asked Questions