Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Customize Warp keyboard shortcuts (keybindings, keymappings) by editing the user's keybindings.yaml file. Use when the user asks to remap a key combination, rebind an action, change a shortcut, or remove a default keybinding (e.g. "change ctrl+space to ctrl+s", "rebind the command palette to cmd+p", "remove the default for X").
.claude/skills/warpdotdev-change-keybinding/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -19% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 51% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 20% | 0% |
Use this skill when the user wants to remap, rebind, or remove a Warp keyboard shortcut.
User customizations live at:
{{keybindings_file_path}}This is the exact path Warp reads at launch — it is platform- and channel-specific (e.g. under ~/.warp*/ on macOS, under XDG config dirs like ~/.config/warp-terminal/ on Linux, and under %LocalAppData% on Windows). Use this path verbatim — do not infer a different one from the user's home directory layout. Create the file (and any missing parent directories) if it does not exist.
A flat YAML map of action_name → key_trigger. Action names contain a colon, so they must be quoted:
yaml"workspace:toggle_ai_assistant": ctrl-s "editor_view:delete_all_left": cmd-shift-A "workspace:toggle_command_palette": none
Triggers use Warp's normalized form — get this exactly right or the binding silently fails to load.
ctrl-alt-shift-cmd-meta-. Cross-platform alias: cmdorctrl- (becomes cmd on macOS, ctrl elsewhere).shift, the letter is lowercase (ctrl-s). With shift, the letter is uppercase (shift-A, never shift-a). Mixing them is invalid.space, enter, escape, tab, backspace, delete, insert, up, down, left, right, home, end, pageup, pagedown, f1–f20, numpadenter. Always lowercase, even with shift (ctrl-shift-space, shift-tab — never ctrl-shift-SPACE). Use the literal word space — not " ".cmd-=, cmd-,, cmdorctrl-/.none. The action becomes unbound.Translate user phrasing into this form: Ctrl+S → ctrl-s, Cmd+Shift+P → cmd-shift-P, Ctrl+Space → ctrl-space.
Defaults are compiled into Warp and are not discoverable from the keybindings file on disk. There is no catalog the agent can consult to map a description or current shortcut to an action name. Pick the right strategy based on how the user described the change:
workspace:show_keybinding_settings, default cmd-ctrl-k on macOS; Settings → Keyboard Shortcuts on other platforms) — they can search by description or current shortcut there and either edit the binding in place or share the canonical namespace:action_name so you can write it.{{keybindings_file_path}} if present. Preserve every existing entry — only add or update the one you're changing.settings.toml which hot-reloads. They can quit with cmd-Q (macOS) or the equivalent on their platform and reopen Warp.Remap an existing custom binding by old trigger:
yaml# before "workspace:toggle_ai_assistant": ctrl-space # after "workspace:toggle_ai_assistant": ctrl-s
Remove a default shortcut:
yaml"workspace:toggle_keybindings_page": none
Shift combined with a special key (note the special key stays lowercase):
yaml"workspace:toggle_ai_assistant": ctrl-shift-space
Cross-platform binding using the cmdorctrl- alias (resolves to cmd on macOS, ctrl elsewhere):
yaml"workspace:toggle_command_palette": cmdorctrl-shift-P
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 7,618 | 10,704 | +41% | 1 | 1 | 0% | 853 | 1,625 | +91% | 0 | 0 | — |
case-02 | fail→fail | 9,543 | 10,933 | +15% | 1 | 1 | 0% | 1,384 | 1,622 | +17% | 0 | 0 | — |
case-03 | fail→fail | 11,310 | 8,857 | -22% | 1 | 1 | 0% | 1,822 | 2,562 | +41% | 0 | 0 | — |
case-04 | fail→pass | 11,524 | 3,939 | -66% | 1 | 1 | 0% | 2,004 | 1,619 | -19% | 0 | 0 | — |
case-05 | pass→pass | 10,220 | 4,410 | -57% | 1 | 1 | 0% | 1,800 | 1,710 | -5% | 0 | 0 | — |
case-06 | fail→pass | 12,403 | 4,758 | -62% | 1 | 1 | 0% | 1,938 | 1,854 | -4% | 0 | 0 | — |
case-07 | pass→pass | 9,031 | 7,982 | -12% | 1 | 1 | 0% | 1,343 | 2,266 | +69% | 0 | 0 | — |
case-08 | pass→pass | 9,724 | 4,372 | -55% | 1 | 1 | 0% | 1,382 | 1,757 | +27% | 0 | 0 | — |
case-09 | fail→pass | 11,117 | 9,576 | -14% | 1 | 1 | 0% | 1,756 | 2,653 | +51% | 0 | 0 | — |
case-10 | fail→pass | 11,037 | 5,054 | -54% | 1 | 1 | 0% | 1,879 | 1,883 | +0% | 0 | 0 | — |
case-11 | pass→pass | 23,702 | 5,206 | -78% | 1 | 1 | 0% | 1,931 | 1,807 | -6% | 0 | 0 | — |
case-12 | pass→pass | 5,611 | 11,958 | +113% | 1 | 1 | 0% | 935 | 1,658 | +77% | 0 | 0 | — |
case-13 | fail→pass | 8,247 | 2,629 | -68% | 1 | 1 | 0% | 1,190 | 1,428 | +20% | 0 | 0 | — |
case-14 | fail→pass | 10,439 | 7,871 | -25% | 1 | 1 | 0% | 1,841 | 2,395 | +30% | 0 | 0 | — |
case-15 | fail→pass | 10,090 | 7,986 | -21% | 1 | 1 | 0% | 1,391 | 2,319 | +67% | 0 | 0 | — |
case-16 | pass→pass | 11,242 | 4,300 | -62% | 1 | 1 | 0% | 1,544 | 1,713 | +11% | 0 | 0 | — |
case-17 | fail→pass | 11,606 | 4,316 | -63% | 1 | 1 | 0% | 1,809 | 1,755 | -3% | 0 | 0 | — |
case-18 | fail→pass | 10,289 | 5,877 | -43% | 1 | 1 | 0% | 1,506 | 1,980 | +31% | 0 | 0 | — |
case-19 | pass→pass | 11,409 | 3,134 | -73% | 1 | 1 | 0% | 1,746 | 1,484 | -15% | 0 | 0 | — |
case-20 | pass→pass | 8,521 | 2,451 | -71% | 1 | 1 | 0% | 1,361 | 1,352 | -1% | 0 | 0 | — |
case-21 | pass→pass | 10,828 | 7,778 | -28% | 1 | 1 | 0% | 1,913 | 2,329 | +22% | 0 | 0 | — |
case-22 | pass→pass | 9,596 | 22,330 | +133% | 1 | 1 | 0% | 1,495 | 2,550 | +71% | 0 | 0 | — |
case-23 | pass→pass | 8,619 | 4,808 | -44% | 1 | 1 | 0% | 1,283 | 1,761 | +37% | 0 | 0 | — |
case-24 | pass→pass | 11,166 | 7,452 | -33% | 1 | 1 | 0% | 1,680 | 2,359 | +40% | 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. 24 cases were attempted, and 22 counted toward the lift figure. The other 2 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 +38 percentage points is the difference between those two pass rates over the 22 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.