Catch list-page bugs before bulk actions hit wrong rows

    1

    Does not run the app, edit code or review styling, performance or accessibility.

    $6

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Catch list-page bugs before bulk actions hit wrong rows

    Example session with this skill installed

    Audit our Users admin page before release: UsersPage.tsx, UsersTable.tsx, BulkBar.tsx and the API handler users.handler.ts. Admins search, filter by status, sort columns and bulk-block users.

    • Read your context and instructions
    • Compiled the catch list-page bugs
    • Generated the document

    List page audit: Users (UsersPage.tsx, UsersTable.tsx, BulkBar.tsx, users.handler.ts)

    Verdict: 2 Critical, 5 High, 3 Medium, 2 Low — fix B4 and B3 before anyone uses "Block".

    State map

    StateLives inWritten byRead byResets when
    pageURL ?pagegoToPageeffect request, Prev buttonnever (not on search/status change)
    searchcomponent state (seeded once from URL ?q)input onChangeeffect requestnever; URL not updated
    statuscomponent stateselect onChangeeffect requestnever; not in URL
    sortBycomponent stateheader clickclient-side sorted onlynever; not sent to server
    selectedcomponent state, row indexescheckboxBulkBar, blockSelectedonly after a bulk action
    rows / totalcomponent stateeffect .thentable, "N users"overwritten in arrival order

    Findings (most severe first)

    [Critical] B4 Selection is stored as row index, and the index points into the client-sorted array

    Evidence: UsersTable.tsx — checked={selected.includes(i)} onChange={() => onToggle(i)}; UsersPage.tsx — const ids = selected.map((i) => sorted[i].id);

    Scenario

    1. Tick Alice and Bob (rows 0 and 1).
    2. Click the "Name" header.
    3. The ticks stay on rows 0 and 1, now different users.
    4. Block → the wrong two users are blocked.
      Fix: store selected ids (Set<string>), toggle with r.id, send [...selected].
      Confidence: confirmed from code
    [Critical] B3 Selection survives search, status and page changes

    Evidence: setSelected([]) is called only inside blockSelected.

    Scenario

    1. Tick rows 3 and 4.
    2. Type "smith".
    3. BulkBar still says "2 selected".
    4. Block → whichever "smith" users sit at positions 3 and 4 are blocked; on a shorter list sorted[i].id throws.
      Fix: clear selection whenever page, search, status or sort changes.
      Confidence: confirmed from code
    [High] A1 Sorting by column sorts only the 25 rows of the current page

    Evidence: const sorted = [...rows].sort(...); sortBy is never sent; server always sorts { createdAt: -1 }.
    Fix: send sort to the server, allowlist it there, drop the client sort.

    [High] C1 Older search response can overwrite a newer one

    Evidence: effect .then(res => setRows(res.rows)) with no cleanup/abort; search changes per keystroke.
    Scenario (slow network): type "jo" then "jon"; the "jo" response lands last → input says "jon", table shows "jo".
    Fix: ignore flag in the effect cleanup, or a keyed query hook; debounce input.

    [High] F2 Total count ignores the search filter

    Evidence: users.handler.ts — page query uses filter, count uses { status: filter.status }.
    Fix: User.countDocuments(filter).

    [High] E2 After "Block", the list does not refetch

    Evidence: goToPage(page) pushes the same ?page=; effect deps [page, search, status] don't change.

    [High] F1 Server sort has no unique tie-breaker

    Fix: .sort({ createdAt: -1, _id: -1 }). Confidence: likely, depends on how often createdAt repeats.

    [Medium] B1 Search/status change keeps the page number → "No users found" on page 8 of 1
    [Medium] G1 A failed request is shown as "No users found"
    [Medium] F3 No upper bound on page size
    [Low] D1 Search read from the URL once; status not in the URL
    [Low] D2 ?page=abc gives NaN on the client

    Questions

    • Is the page meant to be shared by link? If yes, D1 rises to Medium.
    • What does your Mongoose/driver config do with { status: undefined } in the count?

    Not checked

    • Server side of api.blockUsers (scoping of ids).

    Outside this audit

    • req.query.q goes into $regex unescaped — worth a security review.

    UsersPage.tsx

    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

    Why

    Admin tables look fine in a demo with 12 rows. In production, a column sort only sorts the current page, a search response from two keystrokes ago overwrites the new one, and a selection kept by row index makes "Block 2 users" block the wrong two. None of this fails TypeScript or lint.

    What it checks

    The skill first builds a state map (where page, page size, sort, filters, search, selection and totals live, who writes them, when they reset), then runs seven groups of checks:

    • Scope — client-side sort/filter/search on a server-paginated page, totals from page length, "select all" and export that don't match their labels
    • Reset rules — page not reset on filter change, selection surviving filter/sort/page changes, selection keyed by index instead of id
    • Request ordering — late responses overwriting newer ones, query keys missing a parameter, debounced input clobbered by URL sync, bulk actions enabled while placeholder rows are on screen
    • URL state — double sources of truth, unvalidated page/sort values, history spam
    • Bulk actions — what's sent vs what was confirmed, missing refetch/rollback, double-fire, server-side scoping of ids (if the server code is provided)
    • Server pagination — sort without a unique tie-breaker, count and page queries using different filters, uncapped page size
    • States — errors rendered as "No results"

    What you get

    A verdict line, the state map, findings ranked Critical → Low (Critical is reserved for actions on rows the user did not see or confirm), each with evidence, a reproduction a tester can follow, a minimal fix and a confidence label. Then open questions and what was not checked.

    Limits

    Reads files only (optionally read-only git diff/log/show). Does not run the app, edit code or review styling, performance or accessibility. Library notes (TanStack Query v5, Next.js App Router search params, MongoDB/SQL tie-breakers) were checked against official docs in November 2026.

    Reviews list-state correctness only: no styling, accessibility, performance or general security review (serious unrelated issues get a one-line pointer). Does not execute code, so timing bugs are reported with the conditions needed to reproduce them, not reproduced. Server-side checks need the server handler. Not built for virtualised infinite-scroll feeds or spreadsheet-style editable grids.

    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 4 days ago

    Requires

    Git

    Permissions

    File Scopes

    List
    page
    components
    data
    hooks
    and
    API
    handlers
    the
    user
    points
    to
    (read-only)

    read/grep/glob: to open and search the list page, its data hook and server handler. bash is optional and limited to read-only git commands (git diff, git log, git show) to focus on changed code; the skill never writes files, installs packages, runs the app or contacts any network host.

    Listed4 days ago

    What's inside

    Frequently Asked Questions

    No. With only the client it audits the client and lists server-side checks (count/filter mismatch, tie-breakers, page-size caps, id scoping on bulk actions) under 'Not checked'. With the handler it checks those too.

    No. It only reads the files you give it or that it finds in the repo. If a terminal is available it may run read-only git commands (diff, log, show) to focus on changed code. Fixes are short sketches you apply yourself.

    Any React data layer. It has verified notes for TanStack Query v5, Next.js App Router search params (including the Promise-based page prop since Next.js 15) and MongoDB/SQL pagination. Other libraries are reviewed from the code itself, and version assumptions are stated.

    You get the state map, 'No state bugs found in the code I was given', plus any open questions. It is instructed not to pad the report.

    New in Frontend & Web Apps