Install any skill in seconds. Free to start, no credit card required.
Get Started Free →View or modify Warp application settings using the bundled JSON schema for guidance
.claude/skills/warpdotdev-modify-settings/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | -53% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 26% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 64% | 0% |
| case-07 | ✓→✗ | ▼ Worse | -49% | 0% |
| case-08 | ✓→✗ | ▼ Worse | -29% | 0% |
Use this skill when the user wants to view, change, or troubleshoot Warp application settings.
A JSON schema describing all available settings is bundled at:
{{settings_schema_path}}The schema follows JSON Schema draft 2020-12, with settings organized hierarchically under properties. Each setting includes:
description — what the setting controlstype — the value type (string, boolean, integer, etc.)default — the default valueenum or oneOf — valid values, when the setting is constrainedUse grep to do an initial broad search for candidate key names:
shgrep -i "font" {{settings_schema_path}}
Once you have a candidate key name, run the bundled script to get the full dotted path, the setting's properties, and any parent context. This is critical — the schema has multiple sections with similar names (e.g. several input keys), so never assume the nesting from grep output alone.
shpython3 {{skill_dir}}/scripts/find_setting.py {{settings_schema_path}} <key_name>
The output gives you the unambiguous full path (e.g. properties.appearance.properties.input.properties.input_mode) and the setting's full definition including valid values.
The user's settings are stored in a TOML file at:
{{settings_file_path}}Settings use dotted TOML section headers matching the schema hierarchy. Always trace the full nesting path from the schema to the TOML — each intermediate properties key becomes a section level. For example:
A property at properties.appearance.properties.font_size (one level deep) corresponds to:
toml[appearance] font_size = 14
A property at properties.appearance.properties.themes.properties.theme (two levels deep) corresponds to:
toml[appearance.themes] theme = "light"
A common mistake is to stop one level too early — always count the full depth before writing the TOML section header.
If the file does not exist yet, create it. Warp hot-reloads this file, so changes take effect immediately.
grep to identify candidate key names, then run the Python path-tracing script to get the full dotted path and the setting's valid values. Never rely on grep output alone to infer nesting.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 9,819 | 31,113 | +217% | 1 | 1 | 0% | 371 | 906 | +144% | 0 | 0 | — |
case-02 | fail→fail | 36,570 | 13,367 | -63% | 1 | 1 | 0% | 2,173 | 1,067 | -51% | 0 | 0 | — |
case-03 | fail→fail | 5,498 | 13,245 | +141% | 1 | 1 | 0% | 530 | 1,018 | +92% | 0 | 0 | — |
case-04 | pass→pass | 15,876 | 15,051 | -5% | 1 | 1 | 0% | 1,230 | 2,409 | +96% | 0 | 0 | — |
case-05 | pass→pass | 7,944 | 8,422 | +6% | 1 | 1 | 0% | 1,276 | 2,053 | +61% | 0 | 0 | — |
case-06 | pass→pass | 10,900 | 17,109 | +57% | 1 | 1 | 0% | 2,023 | 2,919 | +44% | 0 | 0 | — |
case-07 | pass→fail | 13,864 | 14,244 | +3% | 1 | 1 | 0% | 1,983 | 1,021 | -49% | 0 | 0 | — |
case-08 | pass→fail | 8,285 | 7,356 | -11% | 1 | 1 | 0% | 1,291 | 914 | -29% | 0 | 0 | — |
case-09 | fail→pass | 15,878 | 3,270 | -79% | 1 | 1 | 0% | 2,238 | 1,041 | -53% | 0 | 0 | — |
case-10 | fail→pass | 16,207 | 19,106 | +18% | 1 | 1 | 0% | 1,602 | 2,025 | +26% | 0 | 0 | — |
case-11 | fail→pass | 9,674 | 8,927 | -8% | 1 | 1 | 0% | 1,337 | 2,195 | +64% | 0 | 0 | — |
case-12 | pass→fail | 9,406 | 14,086 | +50% | 1 | 1 | 0% | 1,320 | 1,161 | -12% | 0 | 0 | — |
case-13 | pass→fail | 6,054 | 9,213 | +52% | 1 | 1 | 0% | 910 | 1,119 | +23% | 0 | 0 | — |
case-14 | fail→fail | 9,756 | 8,130 | -17% | 1 | 1 | 0% | 1,223 | 1,056 | -14% | 0 | 0 | — |
case-15 | fail→fail | 10,345 | 16,910 | +63% | 1 | 1 | 0% | 1,342 | 3,458 | +158% | 0 | 0 | — |
case-16 | fail→fail | 11,588 | 8,714 | -25% | 1 | 1 | 0% | 1,859 | 868 | -53% | 0 | 0 | — |
case-17 | fail→fail | 6,988 | 9,792 | +40% | 1 | 1 | 0% | 1,022 | 1,003 | -2% | 0 | 0 | — |
case-18 | fail→fail | 11,000 | 9,553 | -13% | 1 | 1 | 0% | 1,705 | 1,023 | -40% | 0 | 0 | — |
case-19 | fail→fail | 8,891 | 7,700 | -13% | 1 | 1 | 0% | 1,414 | 970 | -31% | 0 | 0 | — |
case-20 | fail→fail | 12,465 | 22,510 | +81% | 1 | 1 | 0% | 1,631 | 1,275 | -22% | 0 | 0 | — |
case-21 | fail→fail | 12,707 | 8,422 | -34% | 1 | 1 | 0% | 1,811 | 925 | -49% | 0 | 0 | — |
case-22 | fail→fail | 10,672 | 9,222 | -14% | 1 | 1 | 0% | 1,707 | 1,000 | -41% | 0 | 0 | — |
case-23 | fail→fail | 27,764 | 25,126 | -10% | 1 | 1 | 0% | 2,020 | 884 | -56% | 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. 23 cases were attempted, and 7 counted toward the lift figure. The other 16 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 -4 percentage points is the difference between those two pass rates over the 7 comparable cases. 4 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.