Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Build or adapt a local harness to drive, inspect, and profile an interactive CLI or TUI without external services. Use for CLI UX checks, startup regressions, memory leaks, hangs, prompt flows, or terminal demos.
.claude/skills/kunanonj-cursor-plugin-teamkit-control-cli/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-20 | ✗→✓ | ▲ Improved | -14% | 0% |
| case-05 | ✓→✗ | ▼ Worse | -30% | 0% |
| case-04 | ✓→✓ | = Same ✓ | 91% | 0% |
| case-06 | ✓→✓ | = Same ✓ | 13% | 0% |
| case-07 | ✓→✓ | = Same ✓ | 30% | 0% |
Use a repeatable local harness to exercise an interactive CLI instead of poking at it manually. First reuse the repo's own test/demo harness if it exists; otherwise assemble a temporary harness from standard local tools.
tmux: managed sessions, capture-pane, send-keys, attach/detach.bashSESSION="cli-harness-$(date +%s)" tmux new-session -d -s "$SESSION" -- <command-under-test> tmux capture-pane -pt "$SESSION" tmux send-keys -t "$SESSION" "help" Enter tmux capture-pane -pt "$SESSION" tmux kill-session -t "$SESSION"
For Node CLIs:
bashNODE_OPTIONS="--inspect=127.0.0.1:0" tmux new-session -d -s "$SESSION" -- <node-cli-command>
Read the terminal output to find the inspector URL, then use Chrome DevTools-compatible tooling if profiling is needed.
Use a PTY script when you need deterministic waits in a repo that does not have tmux or a demo harness. Keep it temporary unless the user asks to add a reusable test.
pythonimport os import pty import select import subprocess import time master_fd, slave_fd = pty.openpty() proc = subprocess.Popen( ["<command>", "<arg>"], stdin=slave_fd, stdout=slave_fd, stderr=slave_fd, close_fds=True, ) os.close(slave_fd) deadline = time.time() + 30 buffer = b"" while time.time() < deadline: ready, _, _ = select.select([master_fd], [], [], 0.25) if not ready: continue chunk = os.read(master_fd, 4096) buffer += chunk if b"<ready text>" in buffer: os.write(master_fd, b"help\n") break print(buffer.decode(errors="replace")) proc.terminate() os.close(master_fd)
If the CLI needs richer terminal control, use pty.fork() or an existing PTY library.
/tmp unless the repo already has a testing/demo harness.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 13,812 | 3,499 | -75% | 1 | 1 | 0% | 2,808 | 1,250 | -55% | 0 | 0 | — |
case-02 | fail→fail | 3,938 | 3,601 | -9% | 1 | 1 | 0% | 243 | 1,237 | +409% | 0 | 0 | — |
case-03 | fail→fail | 12,281 | 3,642 | -70% | 1 | 1 | 0% | 2,503 | 1,229 | -51% | 0 | 0 | — |
case-04 | pass→pass | 5,819 | 6,271 | +8% | 1 | 1 | 0% | 1,205 | 2,296 | +91% | 0 | 0 | — |
case-05 | pass→fail | 7,695 | 3,229 | -58% | 1 | 1 | 0% | 1,825 | 1,271 | -30% | 0 | 0 | — |
case-06 | pass→pass | 6,891 | 2,705 | -61% | 1 | 1 | 0% | 1,403 | 1,590 | +13% | 0 | 0 | — |
case-07 | pass→pass | 6,473 | 3,133 | -52% | 1 | 1 | 0% | 1,225 | 1,591 | +30% | 0 | 0 | — |
case-08 | pass→pass | 4,871 | 2,954 | -39% | 1 | 1 | 0% | 1,027 | 1,622 | +58% | 0 | 0 | — |
case-09 | pass→pass | 10,080 | 9,085 | -10% | 1 | 1 | 0% | 2,099 | 2,995 | +43% | 0 | 0 | — |
case-10 | pass→pass | 13,185 | 8,824 | -33% | 1 | 1 | 0% | 2,209 | 2,498 | +13% | 0 | 0 | — |
case-11 | pass→pass | 8,478 | 2,639 | -69% | 1 | 1 | 0% | 1,351 | 1,473 | +9% | 0 | 0 | — |
case-12 | pass→pass | 15,664 | 10,223 | -35% | 1 | 1 | 0% | 2,653 | 2,668 | +1% | 0 | 0 | — |
case-13 | pass→pass | 2,455 | 2,709 | +10% | 1 | 1 | 0% | 313 | 1,293 | +313% | 0 | 0 | — |
case-14 | pass→pass | 11,086 | 5,736 | -48% | 1 | 1 | 0% | 1,801 | 2,049 | +14% | 0 | 0 | — |
case-15 | pass→pass | 9,691 | 3,087 | -68% | 1 | 1 | 0% | 2,129 | 1,672 | -21% | 0 | 0 | — |
case-16 | pass→pass | 16,546 | 11,686 | -29% | 1 | 1 | 0% | 2,820 | 3,075 | +9% | 0 | 0 | — |
case-17 | pass→pass | 10,771 | 5,738 | -47% | 1 | 1 | 0% | 1,699 | 1,948 | +15% | 0 | 0 | — |
case-18 | pass→pass | 9,321 | 5,151 | -45% | 1 | 1 | 0% | 1,546 | 1,980 | +28% | 0 | 0 | — |
case-19 | pass→pass | 5,876 | 2,568 | -56% | 1 | 1 | 0% | 1,039 | 1,500 | +44% | 0 | 0 | — |
case-20 | fail→pass | 9,494 | 2,832 | -70% | 1 | 1 | 0% | 1,440 | 1,241 | -14% | 0 | 0 | — |
case-21 | pass→pass | 15,191 | 10,523 | -31% | 1 | 1 | 0% | 2,434 | 2,966 | +22% | 0 | 0 | — |
case-22 | pass→pass | 4,313 | 3,063 | -29% | 1 | 1 | 0% | 753 | 1,540 | +105% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 18 counted toward the lift figure. The other 4 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of 0 percentage points is the difference between those two pass rates over the 18 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.