More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Eval Harness Bottleneck Fix

    1

    Make evaluation runs fast enough to run often, without quietly dropping coverage or hiding quarantined cases.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Eval Harness Bottleneck Fix

    Eval Harness Bottleneck Fix

    Example session with this skill installed

    Setup-heavy case 240,000 ms twice, fast case 200 ms with one failure.

    • Read your context and instructions
    • Compiled the eval harness bottleneck

    Verbatim script output on the shipped sample:

    setup-heavy mean=240000ms share=99.9% runs=2 fail_rate=0% 
    fast mean=200ms share=0.1% runs=2 fail_rate=50% FLAKY
    

    Reading: one setup-heavy case owns 99.9 percent of mean runtime, while the fast case is flaky and needs a fix ticket, not speed work. The full worked example in SKILL.md walks through the same method on a larger realistic case, including the limits of the first fix.

    fix ticket, not speed work. The full worked example in SKILL.md

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

    About this skill

    Eval Harness Bottleneck Fix is a step-by-step workflow for test and evaluation harness runtime diagnosis. It ranks causes by measured impact before any fix is attempted, so effort goes to the largest contributor first. It includes diagnostic trees, thresholds to state in advance, decision tables, templates, checklists, and a scoring rubric, plus a runnable script that performs the core ranking or measurement on real input. The worked example in the skill shows the method end to end on a realistic case, including what remains unresolved after the first fix.

    Results depend on the quality and coverage of the input data; missing logs, missing timings, or a biased sample will limit conclusions. Thresholds in the skill are defaults and must be adjusted to the system at hand. The script performs the core measurement only; judgement about causes, guards, and rollout remains with the user.
    Mean runtime from two runs per case is a demonstration only; the skill separates flakiness from slowness across repeated runs and requires quarantined cases to carry a fix ticket and review date.

    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 23 days ago

    Permissions

    Terminal / Shell
    Listed23 days ago
    Updated21 days ago

    What's inside

    Frequently Asked Questions

    No. It diagnoses, ranks, and guides a fix, and its script measures the inputs. Applying a fix stays a deliberate change with a rollback note.

    The inputs listed in the skill: a baseline window, a sample of the failing cases, and whatever timing, log, or label data the workflow ranks. If an input is missing, the skill says to create it first rather than guess.

    It forces ranking by total measured impact, one change at a time, and re-measurement on the same sample shape, with guard metrics checked. A generic checklist rarely requires any of those.

    New in Workflow Automation