picker-verify

name	picker-verify
description	Runs e2e tests for the tmux Claude Code session picker (home/programs/tmux/picker/picker.tsx) in an isolated tmux server sandbox. Use proactively after any change to picker.tsx, picker_e2e_harness.ts, or picker_e2e_test.ts — even if the user doesn't explicitly ask for verification, run at least once to catch regressions before the user tries the picker. Also use when asked to "verify picker", "picker動作確認", "picker e2e", "picker を検証して", "tmux picker のテスト", or "picker-verify".
allowed-tools	Bash(*picker-verify.ts*)

picker-verify

Verify the tmux picker end-to-end against a fresh isolated tmux server. Warms the Deno module cache, runs every scenario in home/programs/tmux/picker/picker_e2e_test.ts, and writes a machine-readable JSON result to stdout.

Quick Start

.claude/skills/picker-verify/picker-verify.ts

Workflow

Step 1: Run the script

.claude/skills/picker-verify/picker-verify.ts

The script is self-contained. It:

1. Runs deno cache home/programs/tmux/picker/picker.tsx to warm npm modules.
2. Runs picker_e2e_test.ts under deno test with the permission scopes the script defines (see the runDeno call and its scope-rationale comments in picker-verify.ts — that file is the source of truth for the exact flags).
3. Parses the test runner output (scenario names + ok/FAILED).
4. Emits a single JSON object on stdout.

Step 2: Read the JSON result

Success example (illustrative — treat ok as the verdict, not any specific number):

{
  "check": "picker-e2e",
  "ok": true,
  "scenarios": {
    "passed": 37,
    "failed": 0,
    "names_failed": []
  },
  "elapsed_ms": 20502,
  "errors": []
}

Failure example:

{
  "check": "picker-e2e",
  "ok": false,
  "scenarios": {
    "passed": 36,
    "failed": 1,
    "names_failed": ["S3: navigation (Down/Up/jk moves the pointer)"]
  },
  "elapsed_ms": 21104,
  "errors": ["see stderr for full test output"]
}

Step 3: Interpret

• ok: true — every scenario passed. Report to user: "picker-verify: all scenarios pass (N/N)."
• ok: false — at least one scenario failed. Share the names_failed list with the user, then read the test's stderr output (the skill prints it under the JSON) to diagnose.
• Exit code follows ok: 0 when all pass, 1 otherwise.

Step 4: Fix and re-run

If failures are in your current change, fix and re-run this skill. Do not claim picker changes are complete while ok: false.

Scope

This skill only exercises:

• picker.tsx's live rendering + key handling inside a spawned tmux pane
• select-window / select-pane side-effects (switch-client is out of scope)
• Escape-driven clean exit

It does NOT cover:

• picker's unit-level pure helpers (those live in picker_test.ts)
• popup-session.sh (tested through direct launch, not popup)
• visual regression / ANSI diff (out of scope by design)

Timing

Runtime varies with the machine, the Deno cache state, and the scenario count. A run that produces no new scenario output for a long stretch indicates a hung scenario.

Override the per-wait timeout via env var if a slow environment trips false timeouts:

PICKER_E2E_TIMEOUT_MS=10000 .claude/skills/picker-verify/picker-verify.ts