Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Focuses on a specific target object to signal task completion. Execute this skill when you have produced the required final object (like a grown banana) and need to formally complete the assigned task. This handles the 'focus on OBJ' action that typically marks successful task execution in the environment.
.claude/skills/majiayu000-task-completion-focus/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -54% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -17% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -23% | 0% |
This skill is the final step in a task execution chain. It is triggered only when the primary objective of a task has been successfully achieved and the target object is present and observable in the environment. Its sole function is to execute the focus on OBJ action on the correct object, which formally signals task completion to the environment.
look around or examine if necessary.focus on <OBJECT> action.0).Scenario: Task was to "grow a banana". After successful cultivation, multiple bananas are visible.
focus on banana on banana tree0 (or the first instance of the target banana).This skill is a low-freedom, terminal action. Its logic is simple and deterministic. The creative and complex work (growing the plant, crafting the item) belongs to other skills. This skill exists solely to "press the final button."
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 12,276 | 7,868 | -36% | 1 | 1 | 0% | 2,093 | 968 | -54% | 0 | 0 | — |
case-02 | fail→pass | 13,959 | 4,032 | -71% | 1 | 1 | 0% | 1,351 | 1,125 | -17% | 0 | 0 | — |
case-03 | fail→pass | 19,191 | 3,849 | -80% | 1 | 1 | 0% | 1,433 | 1,067 | -26% | 0 | 0 | — |
case-04 | pass→pass | 7,308 | 9,266 | +27% | 1 | 1 | 0% | 1,105 | 1,569 | +42% | 0 | 0 | — |
case-05 | pass→pass | 13,452 | 9,035 | -33% | 1 | 1 | 0% | 1,357 | 1,052 | -22% | 0 | 0 | — |
case-06 | pass→pass | 7,908 | 3,866 | -51% | 1 | 1 | 0% | 487 | 1,049 | +115% | 0 | 0 | — |
case-07 | pass→pass | 8,509 | 7,714 | -9% | 1 | 1 | 0% | 1,371 | 800 | -42% | 0 | 0 | — |
case-08 | fail→pass | 12,261 | 8,434 | -31% | 1 | 1 | 0% | 1,055 | 928 | -12% | 0 | 0 | — |
case-09 | fail→pass | 36,425 | 3,115 | -91% | 1 | 1 | 0% | 1,253 | 959 | -23% | 0 | 0 | — |
case-15 | pass→pass | 14,747 | 4,296 | -71% | 1 | 1 | 0% | 1,490 | 747 | -50% | 0 | 0 | — |
case-10 | pass→pass | 16,573 | 8,760 | -47% | 1 | 1 | 0% | 961 | 1,066 | +11% | 0 | 0 | — |
case-11 | fail→pass | 16,633 | 8,016 | -52% | 1 | 1 | 0% | 638 | 914 | +43% | 0 | 0 | — |
case-12 | pass→pass | 30,257 | 7,686 | -75% | 1 | 1 | 0% | 2,080 | 884 | -57% | 0 | 0 | — |
case-13 | fail→pass | 12,564 | 7,318 | -42% | 1 | 1 | 0% | 1,848 | 880 | -52% | 0 | 0 | — |
case-14 | pass→pass | 17,710 | 7,541 | -57% | 1 | 1 | 0% | 2,286 | 813 | -64% | 0 | 0 | — |
case-16 | fail→pass | 16,863 | 7,577 | -55% | 1 | 1 | 0% | 1,848 | 796 | -57% | 0 | 0 | — |
case-17 | pass→pass | 6,520 | 6,749 | +4% | 1 | 1 | 0% | 237 | 700 | +195% | 0 | 0 | — |
case-18 | pass→pass | 12,522 | 3,340 | -73% | 1 | 1 | 0% | 1,836 | 959 | -48% | 0 | 0 | — |
case-19 | fail→pass | 17,628 | 7,056 | -60% | 1 | 1 | 0% | 689 | 770 | +12% | 0 | 0 | — |
case-20 | pass→pass | 13,462 | 7,797 | -42% | 1 | 1 | 0% | 1,009 | 780 | -23% | 0 | 0 | — |
case-21 | fail→pass | 45,385 | 7,902 | -83% | 1 | 1 | 0% | 2,126 | 885 | -58% | 0 | 0 | — |
case-22 | fail→pass | 1,776 | 1,734 | -2% | 1 | 1 | 0% | 212 | 652 | +208% | 0 | 0 | — |
case-23 | fail→pass | 15,834 | 8,054 | -49% | 1 | 1 | 0% | 1,636 | 929 | -43% | 0 | 0 | — |
case-24 | fail→pass | 49,774 | 7,699 | -85% | 1 | 1 | 0% | 1,406 | 855 | -39% | 0 | 0 | — |
case-25 | fail→pass | 17,562 | 2,256 | -87% | 1 | 1 | 0% | 1,985 | 795 | -60% | 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. 25 cases were attempted, and 20 counted toward the lift figure. The other 5 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 +56 percentage points is the difference between those two pass rates over the 20 comparable cases.
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.