Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Plan mode for Kheish — inspect context, write a markdown plan into the active workspace's `.kheish/plans/` directory, and do not execute the work.
.claude/skills/graniet-plan/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-13 | ✗→✓ | ▲ Improved | -66% | 0% |
| case-21 | ✗→✓ | ▲ Improved | -62% | 0% |
| case-04 | ✓→✗ | ▼ Worse | -11% | 0% |
| case-11 | ✓→✗ | ▼ Worse | -57% | 0% |
| case-12 | ✓→✗ | ▼ Worse | -27% | 0% |
This skill is repo-local and stays inactive until explicitly activated.
When the original instructions refer to legacy tool names, use these Kheish mappings:
terminal => bashweb_extract => web_fetch, plus web_search when discovery is neededsearch_files => grep_search and glob_searchbrowser_* tools require a browser-capable surfaced tool or MCP; if none is available, use the closest available surface and say so explicitlyWhen the instructions mention local helper files, resolve them from ${KHEISH_SKILL_DIR}.
Use this skill when the user wants a plan instead of execution.
For this turn, you are planning only.
.kheish/plans/.Write a markdown plan that is concrete and actionable.
Include, when relevant:
If the task is code-related, include exact file paths, likely test targets, and verification steps.
Save the plan with write_file under:
.kheish/plans/YYYY-MM-DD_HHMMSS-<slug>.mdTreat that as relative to the active working directory / backend workspace. Kheish file tools are backend-aware, so using this relative path keeps the plan with the workspace on local, docker, ssh, modal, and daytona backends.
If the runtime provides a specific target path, use that exact path. If not, create a sensible timestamped filename yourself under .kheish/plans/.
/plan, infer the task from the current conversation context.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-17 | fail→fail | 3,796 | 5,460 | +44% | 1 | 1 | 0% | 214 | 821 | +284% | 0 | 0 | — |
case-01 | fail→fail | 2,248 | 12,942 | +476% | 1 | 1 | 0% | 342 | 878 | +157% | 0 | 0 | — |
case-02 | fail→fail | 3,369 | 3,923 | +16% | 1 | 1 | 0% | 214 | 843 | +294% | 0 | 0 | — |
case-03 | fail→fail | 3,639 | 5,203 | +43% | 1 | 1 | 0% | 152 | 801 | +427% | 0 | 0 | — |
case-04 | pass→fail | 5,821 | 5,665 | -3% | 1 | 1 | 0% | 995 | 883 | -11% | 0 | 0 | — |
case-05 | fail→fail | 3,601 | 12,476 | +246% | 1 | 1 | 0% | 508 | 1,743 | +243% | 0 | 0 | — |
case-06 | fail→fail | 2,139 | 6,496 | +204% | 1 | 1 | 0% | 181 | 1,013 | +460% | 0 | 0 | — |
case-07 | fail→fail | 5,782 | 6,462 | +12% | 1 | 1 | 0% | 909 | 920 | +1% | 0 | 0 | — |
case-08 | fail→fail | 2,240 | 4,782 | +113% | 1 | 1 | 0% | 221 | 899 | +307% | 0 | 0 | — |
case-09 | fail→fail | 18,369 | 6,120 | -67% | 1 | 1 | 0% | 3,260 | 951 | -71% | 0 | 0 | — |
case-10 | fail→fail | 21,548 | 6,065 | -72% | 1 | 1 | 0% | 2,712 | 835 | -69% | 0 | 0 | — |
case-11 | pass→fail | 12,641 | 5,568 | -56% | 1 | 1 | 0% | 2,046 | 889 | -57% | 0 | 0 | — |
case-12 | pass→fail | 9,798 | 5,939 | -39% | 1 | 1 | 0% | 1,400 | 1,016 | -27% | 0 | 0 | — |
case-13 | fail→pass | 12,527 | 1,378 | -89% | 1 | 1 | 0% | 2,161 | 729 | -66% | 0 | 0 | — |
case-14 | pass→fail | 2,034 | 4,469 | +120% | 1 | 1 | 0% | 249 | 760 | +205% | 0 | 0 | — |
case-15 | fail→fail | 10,439 | 4,298 | -59% | 1 | 1 | 0% | 1,544 | 729 | -53% | 0 | 0 | — |
case-16 | pass→fail | 15,738 | 4,952 | -69% | 1 | 1 | 0% | 2,553 | 743 | -71% | 0 | 0 | — |
case-18 | fail→fail | 19,283 | 4,255 | -78% | 1 | 1 | 0% | 3,029 | 713 | -76% | 0 | 0 | — |
case-19 | pass→fail | 19,158 | 3,633 | -81% | 1 | 1 | 0% | 3,061 | 781 | -74% | 0 | 0 | — |
case-20 | pass→fail | 4,438 | 5,632 | +27% | 1 | 1 | 0% | 632 | 949 | +50% | 0 | 0 | — |
case-21 | fail→pass | 13,275 | 1,468 | -89% | 1 | 1 | 0% | 2,030 | 770 | -62% | 0 | 0 | — |
case-22 | pass→pass | 6,362 | 2,390 | -62% | 1 | 1 | 0% | 938 | 938 | 0% | 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 3 counted toward the lift figure. The other 19 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 3 comparable cases. 7 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.