Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Structure the user's chosen direction into a formal goal tree using KAOS-style AND/OR decomposition. Validate feasibility against ActorProfile and ObstacleReport. Use after obstacle-analysis confirms the direction is viable.
.claude/skills/yogsoth-ai-goal-decomposition/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -66% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -53% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -31% | 0% |
KAOS-style recursive goal decomposition into a validated GoalTree.
| SOP | Purpose | Execution | |-----|---------|-----------| | formulate-top-goal | Formulate the top-level goal statement with user | dialogue | | and-or-decompose | Recursively decompose goal into AND/OR sub-goals | subagent | | validate-leaves | Validate leaf goals are actionable and measurable | subagent | | feasibility-check | Check leaf feasibility against ActorProfile + ObstacleReport | subagent | | ask-decomposition-validation | Present GoalTree to user, get confirmation + priorities | dialogue |
GoalTree + user-confirmed priorities
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | and-or-decompose | KAOS-style recursive goal decomposition. AND decomposition for sub-goals that must ALL be satisfied. OR decomposition for alternative paths where any one suffices. Produces a GoalTree (DAG structure). | | ask-decomposition-validation | Present the GoalTree to the user for confirmation. Ask about reasonableness, missing elements, and priority ordering among sub-goals. | | feasibility-check | Cross-reference the GoalTree against ActorProfile (capabilities), ObstacleReport (known barriers), and timeline (deadline feasibility). Identify infeasible paths and suggest OR alternatives. | | formulate-top-goal | Express the user's chosen research direction as a formal goal statement in the format: 'Achieve what], such that effect], under constraints]'. Confirm with user before proceeding to decomposition. | | validate-leaves | Quality check on leaf nodes of the GoalTree. Verify each leaf is specific enough, actionable, and that the set of leaves fully covers the top goal. Flag issues for further decomposition. |
<!-- END available-tables (generated) -->
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | fail→pass | 11,660 | 5,143 | -56% | 1 | 1 | 0% | 1,741 | 1,282 | -26% | 0 | 0 | — |
case-05 | fail→pass | 13,497 | 4,508 | -67% | 1 | 1 | 0% | 2,475 | 1,313 | -47% | 0 | 0 | — |
case-01 | fail→pass | 22,216 | 7,848 | -65% | 1 | 1 | 0% | 3,722 | 1,282 | -66% | 0 | 0 | — |
case-02 | fail→pass | 18,080 | 5,436 | -70% | 1 | 1 | 0% | 3,211 | 1,502 | -53% | 0 | 0 | — |
case-03 | fail→fail | 21,942 | 3,852 | -82% | 1 | 1 | 0% | 3,381 | 1,124 | -67% | 0 | 0 | — |
case-06 | pass→pass | 9,534 | 8,451 | -11% | 1 | 1 | 0% | 1,723 | 2,017 | +17% | 0 | 0 | — |
case-07 | fail→pass | 14,482 | 6,964 | -52% | 1 | 1 | 0% | 2,341 | 1,611 | -31% | 0 | 0 | — |
case-08 | pass→pass | 15,221 | 3,779 | -75% | 1 | 1 | 0% | 2,295 | 1,144 | -50% | 0 | 0 | — |
case-09 | pass→pass | 17,023 | 4,971 | -71% | 1 | 1 | 0% | 2,488 | 1,259 | -49% | 0 | 0 | — |
case-10 | pass→pass | 6,861 | 4,111 | -40% | 1 | 1 | 0% | 1,054 | 1,178 | +12% | 0 | 0 | — |
case-11 | pass→pass | 11,060 | 4,255 | -62% | 1 | 1 | 0% | 1,643 | 1,151 | -30% | 0 | 0 | — |
case-12 | fail→fail | 11,282 | 3,351 | -70% | 1 | 1 | 0% | 1,788 | 1,058 | -41% | 0 | 0 | — |
case-13 | fail→pass | 11,713 | 2,360 | -80% | 1 | 1 | 0% | 1,746 | 770 | -56% | 0 | 0 | — |
case-14 | pass→pass | 12,152 | 5,305 | -56% | 1 | 1 | 0% | 1,957 | 1,308 | -33% | 0 | 0 | — |
case-15 | fail→pass | 9,982 | 7,016 | -30% | 1 | 1 | 0% | 1,685 | 1,493 | -11% | 0 | 0 | — |
case-21 | pass→fail | 18,526 | 5,074 | -73% | 1 | 1 | 0% | 3,161 | 1,336 | -58% | 0 | 0 | — |
case-16 | pass→pass | 12,972 | 6,298 | -51% | 1 | 1 | 0% | 2,107 | 1,518 | -28% | 0 | 0 | — |
case-17 | pass→pass | 7,630 | 1,341 | -82% | 1 | 1 | 0% | 1,205 | 638 | -47% | 0 | 0 | — |
case-18 | pass→pass | 12,995 | 1,640 | -87% | 1 | 1 | 0% | 2,171 | 771 | -64% | 0 | 0 | — |
case-19 | pass→fail | 14,566 | 6,525 | -55% | 1 | 1 | 0% | 3,043 | 1,652 | -46% | 0 | 0 | — |
case-20 | pass→pass | 16,699 | 22,676 | +36% | 1 | 1 | 0% | 3,520 | 5,070 | +44% | 0 | 0 | — |
case-22 | pass→pass | 14,569 | 8,375 | -43% | 1 | 1 | 0% | 2,152 | 1,799 | -16% | 0 | 0 | — |
case-23 | pass→pass | 16,476 | 8,231 | -50% | 1 | 1 | 0% | 2,435 | 1,609 | -34% | 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. The headline lift of +22 percentage points is the difference between those two pass rates over the 23 comparable cases. 2 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.