More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Universal Troubleshooter: Diagnose It, Fix It, Prove It's Fixed

    1

    A fixed diagnose-fix-verify routine for any technical failure: reproduce it, read the real error, rank testable causes, change one thing at a time, and prove the fix holds.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Universal Troubleshooter: Diagnose It, Fix It, Prove It's Fixed

    Universal Troubleshooter: Diagnose It, Fix It, Prove It's Fixed

    Example session with this skill installed

    My note-taking app has crashed on launch since yesterday's update. It closes before showing my notes, with no error message, about half the time I open it. I think the update broke it. Can you find the cause and fix it without losing my notes?

    • Read your context and instructions
    • Compiled the universal troubleshooter

    DIAGNOSIS + FIX LOG — Note app crashes on launch — state: current install, after yesterday's update.

    Step 1 — Reproduce: launched the app five times. It closed before showing the note list on 3 of 5 launches, with no message. Intermittent, so frequency is part of the evidence.

    Step 2 — Read the error: the app log's last line before each crash names the thumbnail cache, and one entry stops mid-read. The message names a thing — the cache — so the work starts there.

    Step 3 — Ranked explanations: (a) a corrupt thumbnail cache file; (b) the update broke the launcher; (c) the phone ran out of storage. (a) ranks first: the log points at the cache and the failure began after the update rebuilt it.

    Step 4 — Checks: storage shows plenty free, which rules out (c). The cache folder lists one zero-byte thumbnail file dated yesterday. Renaming the cache folder lets the app launch, which rules out (b) — the launcher is fine.

    Step 5 — Fix: deleted only the one corrupt thumbnail file, nothing else, and restored the folder name.

    Step 6 — Verify: launched the app five times — five clean launches, note list loads every time, a note with images opens, and export still works. Original failure gone, nothing nearby broke. Diagnosis level 4, with the five-run evidence attached.

    Follow-up: the app should skip a bad thumbnail instead of crashing on it; that stays open, and the log entry records it.

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

    About this skill

    When something fails and one error message does not explain it, this skill gives you a fixed order of operations that works across domains: wrong dependencies, config conflicts, broken installs, API failures, database connection problems, runtime errors, and memory or performance faults. It starts by reproducing the failure on purpose in its smallest form, then reading the actual error and output in full instead of guessing from memory. You list every explanation you can think of, keep only the ones that predict an observable check, and rank them by likelihood and by how cheap the check is — then you test the cheapest discriminating check first. Fixes go in one change at a time, so when the failure clears you know which change cleared it, and workarounds are labeled as workarounds instead of recorded as repairs. Verification means running the original reproduction exactly as written down, plus the surrounding checks, so a fix that breaks something else is caught, not celebrated. The skill includes a seven-row explanation ranking table, a copy-ready fix log template, a six-level diagnosis rubric from guess to fixed-verified-guarded, a complete worked example about a bakery ordering service whose database stopped after an overnight update, eight edge cases including failures that only happen under load or on one machine, seven anti-patterns, and an eight-item pre-delivery quality gate. The fix log is the point that outlasts the session: the next person who meets the same failure starts from your entry, not from zero.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Testing & Debugging.

    • It disciplines diagnosis; it does not run your commands, builds, or checks for you, and it cannot diagnose a failure you cannot reproduce or observe at all — in that case it tells you what evidence to gather first.
    • It does not judge whether the feature itself was the right thing to build; a perfectly diagnosed wrong feature still needs a requirements conversation.
    • Diagnosis levels depend on honest reporting of what ran and when; the skill gives you the language to be precise, not a way to detect a misreported run.

    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 buying a skill.

    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 3 days ago

    Permissions

    No special permissions declared or detected

    Listed3 days ago

    What's inside

    Frequently Asked Questions

    Yes, when you can reproduce the failure and read some output from it — a status light, a message, a log. The steps are the same: reproduce, read, rank, test one change at a time, verify, log. It cannot help with a failure you cannot observe at all.

    Then do not start fixing. The skill tells you to narrow the difference between the failing situation and yours — different machine, version, data, or load — because guessing at a failure you cannot see is how you break the working copy too.

    It is slower than changing five things and hoping. It is faster than changing five things, watching the failure clear, and then having no idea which change did it — or which of the other four broke something else. One change at a time is what makes the result mean something.

    New in Testing & Debugging