Skip to content
claudemods

Verify Board Claude Code mod: Claude Code verification that shows when test results go stale

Records your test, lint and typecheck runs and marks each PASS, FAIL, STALE or UNKNOWN against the current git state.

Permissions

  • Reads filesCan read files and list directories on your machine ($.fs.read / list / stat / exists).
  • Runs commandsCan run shell programs on your machine ($.process.run / spawn). Check the source for exactly which ones.
  • Intercepts tool callsSees every tool call Claude makes (tool.call / tool.check) and could block or change it.
  • Changes the UIDraws into the Claude Code interface: bands, panes, status text, toasts (ui.render / $.ui.*).
  • Adds commands/toolsRegisters new slash commands or tools that you or Claude can invoke ($.command.register / $.tool.register).
  • Stores dataKeeps data between sessions in the plugin's private store ($.store.*).
tested with
v2.1.310
last tested
requires
Claude Code ≥ 2.1.287
surfaces
none

Install inside Claude Code:

/plugin install verify-board --marketplace YeonwooSung/my-claude-code-mods
All install options ↓

What does this mod do?

Verify Board is a community mod by YeonwooSung (YeonwooSung), part of the my-claude-code-mods collection. It adds Claude Code verification that answers a narrow question: "the tests passed, but do they still pass for the code as it is now?" When Claude runs npm test and then edits a file, the earlier green result no longer describes your working tree. Verify Board remembers that.

It watches the Bash commands Claude Code runs. When one matches a short list of check commands (tests, typecheck, lint), it stores the command and its outcome together with a fingerprint of the repository at that moment. Type /verify and it prints each recorded check with a status:

  • PASS: the check succeeded and the repository still matches the fingerprint.
  • FAIL: the check failed.
  • STALE: it passed, but files changed afterwards, or the command changed files while it ran.
  • UNKNOWN: the evidence is incomplete, such as a backgrounded or truncated command.

The report is plain text returned to the command. There is no panel or status line.

Install

Use any of the methods in the install block; they install from the author's own marketplace. Reload plugins afterwards. The mod needs a git repository to say anything useful, because the fingerprints come from git.

$ install verify-board

Requires Claude Code ≥ 2.1.287

Installs from the author's own marketplace: YeonwooSung/my-claude-code-mods

  1. 1.One line, inside Claude Code

    /plugin install verify-board --marketplace YeonwooSung/my-claude-code-mods

    Paste into a running session (v2.1.275+). Claude Code asks to add the marketplace first.

  2. 2.From your shell

    claude plugin marketplace add YeonwooSung/my-claude-code-mods
    claude plugin install verify-board@my-claude-code-mods

    Then run /reload-plugins in any session that is already open.

  3. 3.Try it without installing

    git clone https://github.com/YeonwooSung/my-claude-code-mods
    claude --plugin-dir ./my-claude-code-mods/verify-board

    Loads the mod for one session only. Nothing is added to your settings.

How it works

The module hooks/register.ts registers /verify on session.start, answers command.run for that one command, and wraps every tool.call. For Bash calls it takes a repository fingerprint before and after the call, then hands the result to the observer. This is from register.ts:

// next runs outside observation catch: an observer never retries a tool or hides its exception.
const result = await next(e)
try {
  const now = await Promise.all([$.session.cwd(), $.session.id()])
  if (!expected || expected[0] !== now[0] || expected[1] !== now[1]) return result
  await transact($, async (model, io) => {
    const after = before !== undefined ? await io.revision() : undefined
    await model.observe?.({ tool: e.tool, input: object(e), result, before, after }, io)
  }, expected)
} catch { try { $.ui.toast(`${name}: observation not saved`) } catch { /* Noninteractive host. */ } }
return result

The fingerprint in core/runtime.ts is a hash of HEAD, the unstaged and staged binary diffs, git status, and the contents of up to 200 untracked files (2 MiB in total). If a limit is exceeded, a symlink is involved or a submodule is dirty, the fingerprint is undefined and the result stays UNKNOWN. outcome() reads the exit code, the isError flag, and failure lines from common test runners. Records are keyed by working directory and session id and kept to 200 entries.

Customize it

There is no configuration. The commands are /verify or /verify list for the report, /verify help for a one-line summary, and /verify reset to clear the records for the current folder and session. The list of recognized check commands is the isCheck regular expression in core/runtime.ts; edit it in a local copy if you use other tools.

Permissions & safety

We read every file the plugin can execute before running it: plugin.json, hooks/hooks.json, hooks/register.ts, hooks/core/feature.ts, hooks/core/redact.ts and hooks/core/runtime.ts, plus the three test files. There is no package.json, install script, binary, minified code or dynamic import. Nothing in the plugin was left unread. Findings:

  • Network: none. We found no fetch, HTTP client, socket or telemetry.
  • Processes: $.process.run is limited by the code to git, with a 10-second timeout and an argument array (no shell): rev-parse --show-toplevel, rev-parse --verify HEAD, diff --no-ext-diff --no-textconv --binary (plain and --cached), status --porcelain=v1 -z and ls-files --others. Git runs inside your repository, so it honors your git configuration.
  • File reads: $.fs.read and $.fs.stat read untracked files in the repository to hash them. The contents are hashed, not stored.
  • File writes and settings: none outside the plugin store. It does not touch ~/.claude, settings.json, shell files or the project.
  • Stored data: the check command text (secrets masked on a best-effort basis, as the README says), its outcome and a revision hash, in the plugin store, scoped to folder and session. Treat the masking as a safety net, not a guarantee.
  • Tool calls: the tool.call hook forwards every call once and returns its result unchanged. It never blocks, rewrites or retries a call. It does add work: for every Bash call it computes the fingerprint twice, which means several git commands each time, even for commands that are not checks. On a very large repository this can slow Bash calls. claude plugin validate notes both hooks as gating hooks without .catch; in the code, the work before next is wrapped in try or .catch, and the work after it is wrapped in try.
  • Model influence: none. It injects no text into prompts or results; /verify returns text to you.
  • Daemons, secrets, external requirements: none. It reads no environment variables or credentials, and needs only git.
  • License: Apache-2.0, with a LICENSE file in the repository root.

claude plugin validate passed with no errors or warnings. The author's 22 tests passed on Claude Code 2.1.310.

Reads files
Can read files and list directories on your machine ($.fs.read / list / stat / exists).
Runs commands
Can run shell programs on your machine ($.process.run / spawn). Check the source for exactly which ones.
Intercepts tool calls
Sees every tool call Claude makes (tool.call / tool.check) and could block or change it.
Changes the UI
Draws into the Claude Code interface: bands, panes, status text, toasts (ui.render / $.ui.*).
Adds commands/tools
Registers new slash commands or tools that you or Claude can invoke ($.command.register / $.tool.register).
Stores data
Keeps data between sessions in the plugin's private store ($.store.*).

Compatibility & troubleshooting

We loaded it on Claude Code 2.1.310 on macOS; the author says they tested with 2.1.302. It needs function hooks. If /verify shows nothing:

  • "No checks observed" means no matching Bash command has run in this folder and session. Run npm test (or another listed command) and try again.
  • A result stays UNKNOWN if the command was backgrounded, truncated, denied or timed out, or if the folder is not a git repository.
  • A passing check turns STALE as soon as any tracked or untracked file changes, which is intended.
  • After installing into an open session, run /reload-plugins.

FAQ

What do PASS, FAIL, STALE and UNKNOWN mean in Verify Board?
PASS means the check succeeded and the repository content is the same as when it ran. FAIL means it failed. STALE means it passed but files changed afterwards, or the command itself changed files. UNKNOWN means the evidence is incomplete, for example a backgrounded, truncated or denied command, or a folder that is not a git repository.
Which commands does it count as a check?
Only a conservative list matched against the whole Bash command, such as npm, pnpm or yarn test, typecheck and lint, npx tsc, vitest, jest or eslint, pytest, cargo test, check or clippy, and go test. Commands with ;, &, |, < or > are ignored, as are --help, --version and --list style runs.
Does Verify Board run my tests for me?
No. It watches Bash tool calls that Claude or you already made and records the outcome. It never runs a check itself and never blocks or repeats a tool call.
Does it send my code or commands anywhere?
No network code exists in the plugin. It runs only local git commands, and the records it keeps are stored in the plugin's own store with secrets masked on a best-effort basis.
How do I clear the records?
Run /verify reset. It clears only this mod's records for the current folder and session.
Tool Call Counter running in Claude CodeFeatured

Tool Call Counter

Counts Claude's tool calls per turn and session, shows the count by the spinner, and adds /tally for a breakdown.

  • spinner
  • transcript
  • status
  • Intercepts tool calls
  • Changes the UI
  • Adds commands/tools

tested with v2.1.310

>_ show-your-work

Show Your Work

Claude plays screenshots and videos of its work inline in Ghostty, with before/after compare and QA checks.

  • transcript
  • pane
  • Reads files
  • Runs commands
  • Network
  • Intercepts tool calls
  • Reads your prompts
  • Changes the UI
  • Adds commands/tools
  • Stores data

tested with v2.1.296

More in Timers, todos, standups, Slash commands.