Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Scope iTerm automation to the caller tab and pane. Use when working in iTerm and the task involves the current tab, caller pane, non-disruptive splits, preserving focus, routing commands to the right pane, or avoiding interference with another visible iTerm pane.
.claude/skills/antonio-mello-ai-iterm-workspace/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-19 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -23% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -32% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -65% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -2% | 0% |
Use this skill to keep iTerm automation scoped to the pane and tab that invoked the agent. iTerm does not provide a cmux-style workspace API, so the practical workspace unit is the current iTerm tab.
Anchor on the caller pane first. If the agent process has no TTY, use the current iTerm session as a fallback and say so before making visible changes.
bashiterm-control identify iterm-control list
identify reports:
caller_tty: local terminal TTY if available.resolved_caller.resolution: caller-tty, caller-fallback-current, or another explicit resolution.current: iTerm's current session.Build layout additively from the caller/current pane. Prefer commands that create the pane already running the intended command:
bashiterm-control split --target caller --direction right --command "zsh -lc 'cd /repo && exec codex'" iterm-control grid --target caller --rows 2 --cols 2 --command "zsh -lc 'cd /repo && exec codex'"
Avoid focus-changing AppleScript unless the user explicitly asks to move focus. When a target is ambiguous, stop and ask for the pane/session id or use a badge or screenshot to verify.
Use this order for target selection:
session:<id> from a fresh list or identify.tty:<path> from a fresh list or identify.caller when identify resolved by caller-tty.current only when the task is explicitly about the currently focused iTerm pane.Do not send commands to another pane just because it is visually near the caller. iTerm pane geometry is not exposed reliably through AppleScript.
caller-fallback-current as focus-based and user-visible.list, screenshot, or visible output.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-21 | pass→pass | 11,420 | 10,265 | -10% | 1 | 1 | 0% | 2,415 | 2,721 | +13% | 0 | 0 | — |
case-01 | fail→fail | 12,366 | 5,597 | -55% | 1 | 1 | 0% | 2,530 | 756 | -70% | 0 | 0 | — |
case-19 | fail→pass | 8,730 | 2,218 | -75% | 1 | 1 | 0% | 1,633 | 968 | -41% | 0 | 0 | — |
case-02 | fail→fail | 13,127 | 3,511 | -73% | 1 | 1 | 0% | 2,776 | 737 | -73% | 0 | 0 | — |
case-03 | fail→fail | 4,661 | 3,650 | -22% | 1 | 1 | 0% | 638 | 720 | +13% | 0 | 0 | — |
case-04 | pass→pass | 5,150 | 2,225 | -57% | 1 | 1 | 0% | 936 | 998 | +7% | 0 | 0 | — |
case-05 | pass→pass | 10,607 | 1,812 | -83% | 1 | 1 | 0% | 2,005 | 875 | -56% | 0 | 0 | — |
case-06 | fail→pass | 7,300 | 1,995 | -73% | 1 | 1 | 0% | 1,193 | 913 | -23% | 0 | 0 | — |
case-07 | pass→fail | 11,987 | 3,536 | -71% | 1 | 1 | 0% | 2,137 | 1,182 | -45% | 0 | 0 | — |
case-08 | fail→pass | 8,191 | 2,450 | -70% | 1 | 1 | 0% | 1,474 | 1,001 | -32% | 0 | 0 | — |
case-09 | fail→fail | 12,786 | 4,030 | -68% | 1 | 1 | 0% | 2,488 | 732 | -71% | 0 | 0 | — |
case-10 | pass→pass | 9,295 | 4,101 | -56% | 1 | 1 | 0% | 1,788 | 1,206 | -33% | 0 | 0 | — |
case-11 | fail→fail | 4,705 | 4,766 | +1% | 1 | 1 | 0% | 905 | 1,432 | +58% | 0 | 0 | — |
case-12 | fail→fail | 4,609 | 5,767 | +25% | 1 | 1 | 0% | 973 | 976 | +0% | 0 | 0 | — |
case-13 | fail→fail | 8,501 | 5,542 | -35% | 1 | 1 | 0% | 1,582 | 954 | -40% | 0 | 0 | — |
case-14 | fail→pass | 16,115 | 2,858 | -82% | 1 | 1 | 0% | 2,981 | 1,037 | -65% | 0 | 0 | — |
case-20 | pass→pass | 6,347 | 4,365 | -31% | 1 | 1 | 0% | 1,316 | 1,404 | +7% | 0 | 0 | — |
case-15 | fail→pass | 7,560 | 3,848 | -49% | 1 | 1 | 0% | 1,226 | 1,198 | -2% | 0 | 0 | — |
case-16 | pass→pass | 9,151 | 2,539 | -72% | 1 | 1 | 0% | 1,579 | 934 | -41% | 0 | 0 | — |
case-17 | pass→pass | 6,798 | 4,008 | -41% | 1 | 1 | 0% | 1,194 | 1,300 | +9% | 0 | 0 | — |
case-18 | fail→pass | 12,087 | 1,988 | -84% | 1 | 1 | 0% | 2,404 | 914 | -62% | 0 | 0 | — |
case-22 | pass→pass | 9,762 | 10,417 | +7% | 1 | 1 | 0% | 1,827 | 2,654 | +45% | 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 16 counted toward the lift figure. The other 6 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 +23 percentage points is the difference between those two pass rates over the 16 comparable cases. 1 case got worse with the skill loaded, and it is 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.