More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Expo Skill Feedback: Report It Right, Get It Fixed

    1

    Report problems with Expo skills or Expo itself, nominate agent failures as eval candidates, and check opt-in telemetry — three independent mechanisms, each with its own rules.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Expo Skill Feedback: Report It Right, Get It Fixed

    Expo Skill Feedback: Report It Right, Get It Fixed

    Example session with this skill installed

    An agent just spent six attempts getting a formSheet detent right on SDK 56 and I fixed the option name by hand. Turn this into a proper Expo report for me.

    • Read your context and instructions
    • Compiled the expo feedback

    FormSheet failure — report produced with the skill

    Mechanism chosen: eval candidate. The task is solvable and verifiable, the failure was specific, and I took over — that is the signal Expo turns into evals. Ordinary feedback would have buried it.

    Submission drafted and shown to me for approval before sending:

    • Task: Present a form sheet at half detent with a grabber in an Expo Router app on SDK 56.
    • Expected: The sheet opens at half detent; the grabber is visible; dragging to full detent works.
    • Actual: After six attempts the sheet opened full-screen with no grabber; the option name the agent used does not exist on SDK 56.
    • Wrong approach: The agent used the pre-SDK-56 detent option name, hallucinated from older documentation, and kept retrying variations instead of checking the current options.
    • Evidence: Six attempts in one session; I corrected the option name by hand and the sheet then behaved as expected.

    Category evals, subject expo-router, submitted from the app directory so the CLI attaches SDK and package versions. No product context, no pasted code, no secrets anywhere in the text. Rubric score: 9/10 — one point held back because the attempt count is exact but the model name was not recorded; it is noted as unverified rather than guessed.

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

    About this skill

    Expo improves its skills from field reports — but only from reports that are short, actionable, and free of anything that should never leave the machine. This skill separates the three mechanisms that get confused: ordinary feedback (a skill was confusing, broken, or missing context; the CLI or docs fell short), eval-candidate nomination (an agent failed a real task despite genuine effort), and usage telemetry (off by default, independent of feedback, never enabled without an explicit request).

    You get the category/subject cheat sheet for every report type, the rule that no report ever carries secrets, source code, personal data, long prompts, or stack traces, and the eval-candidate bar: solvable, verifiable, specific, one per session, exact text shown to the user and approved before sending, product context excluded. The resend rule is strict — one sanctioned retry for the known evals rejection, none for anything else. A worked example turns a six-attempt formSheet failure into a submission a maintainer can act on in one reading, with a scoring rubric gating every send at 8/10.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Mobile Development.

    • It cannot verify that a submitted report was received or acted on — feedback is a one-way channel.
    • Eval candidates need a user present to approve the exact text; headless or CI runs cannot submit them.
    • It does not file feature requests as bugs — mislabelled reports get the wrong triage, and the skill says to label them honestly instead.
    • The telemetry script path depends on the installed skill location; the environment-variable equivalents are the fallback.

    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

    Software
    submit-expo-feedbackNode.jsexpo

    Permissions

    Terminal / Shell
    Listed3 days ago

    What's inside

    Frequently Asked Questions

    No. The two mechanisms are independent, and telemetry stays off by default. Anyone who tells you feedback requires telemetry is wrong — this skill keeps them separate on purpose.

    Three things: solvable (it was eventually done or is clearly doable), verifiable (success is observable), and specific (one particular mistake, not a vague struggle). Plus user approval of the exact text, and only one per session.

    Only if it rejected the `evals` category — then resend once with `--category unknown` and prefix the subject with `eval-candidate: `. Any other error: do not resend. A failed submission hammered into the queue becomes duplicate reports.

    New in Mobile Development