Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Goal-Driven Requirement Refinement Engine for Research. Crystallize a user's fuzzy research intent into a North Star statement and structured ResearchBrief through adaptive dialogue and on-demand investigation.
.claude/skills/yogsoth-ai-north-star-crystallization/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 2% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -46% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -14% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 10% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 14% | 0% |
| Signal | Route to | |--------|----------| | No direction at all ("I want to publish but don't know what") | cold-start | | Has a general direction but not specific ("I'm interested in LLM reasoning") | warm-start | | Has a specific topic/problem ("I want to improve CoT faithfulness") | hot-start |
| Strategy | Purpose | |----------|---------| | cold-start | Full crystallization from zero — actor profiling through synthesis | | warm-start | Simplified flow for users with a general direction | | hot-start | Rapid crystallization for users with a specific topic |
| Tactic | Purpose | |--------|---------| | actor-profiling | Understand user's background, expertise, resources | | landscape-reconnaissance | Broad survey of potential research areas | | direction-narrowing | Focus from broad area to specific problem | | obstacle-analysis | Identify and assess obstacles to research goals | | goal-decomposition | Break down research goals into sub-goals | | north-star-synthesis | Converge into North Star + ResearchBrief |
Dialogue + investigation operations. See individual SOP directories for details.
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| Strategy | When to use | | --- | --- | | cold-start | Full crystallization strategy for users who have no research direction at all. Covers actor profiling, landscape reconnaissance, direction narrowing, obstacle analysis, goal decomposition, and north-star synthesis. Use when the user's first message reveals zero specificity about what they want to research. | | hot-start | Minimal crystallization strategy for users who already have a specific research topic or problem (e.g., "I want to improve CoT faithfulness in LLMs") and need structuring into a formal North Star. Heavily simplifies or skips exploration tactics, focusing on obstacle analysis, goal decomposition, and synthesis. Use when the user's first message reveals a specific, actionable research direction. | | warm-start | Simplified crystallization strategy for users who have a general research direction (e.g., "I'm interested in LLM reasoning") but lack specificity. Simplifies actor profiling and landscape reconnaissance, then proceeds through direction narrowing, obstacle analysis, goal decomposition, and north-star synthesis. Use when the user's first message reveals a general area but not a specific problem. |
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | context-checkpoint | Append research process and results to the current Phase's context file. Covers both process and results with genuine substance. Use this skill at plan-designated checkpoint points — typically after each strategy completes or at key decision nodes within a research Phase. | | context-init | Create a new context file for a research Phase. Called once at Phase start to initialize the file that subsequent context-checkpoint calls will append to. Use this skill whenever a new research Phase begins and a fresh context file is needed. |
<!-- 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-07 | fail→pass | 27,962 | 28,800 | +3% | 1 | 1 | 0% | 1,611 | 1,649 | +2% | 0 | 0 | — |
case-01 | fail→fail | 36,817 | 17,999 | -51% | 1 | 1 | 0% | 4,300 | 2,620 | -39% | 0 | 0 | — |
case-02 | fail→pass | 36,464 | 14,131 | -61% | 1 | 1 | 0% | 3,933 | 2,105 | -46% | 0 | 0 | — |
case-03 | fail→pass | 25,785 | 15,487 | -40% | 1 | 1 | 0% | 2,607 | 2,238 | -14% | 0 | 0 | — |
case-04 | pass→pass | 45,967 | 12,621 | -73% | 1 | 1 | 0% | 2,744 | 1,957 | -29% | 0 | 0 | — |
case-05 | pass→pass | 16,608 | 8,543 | -49% | 1 | 1 | 0% | 1,771 | 1,261 | -29% | 0 | 0 | — |
case-06 | fail→pass | 27,466 | 6,945 | -75% | 1 | 1 | 0% | 905 | 991 | +10% | 0 | 0 | — |
case-08 | pass→pass | 32,958 | 22,321 | -32% | 1 | 1 | 0% | 1,718 | 1,626 | -5% | 0 | 0 | — |
case-09 | pass→pass | 27,044 | 24,983 | -8% | 1 | 1 | 0% | 1,365 | 1,448 | +6% | 0 | 0 | — |
case-10 | fail→pass | 34,815 | 13,377 | -62% | 1 | 1 | 0% | 954 | 1,089 | +14% | 0 | 0 | — |
case-11 | fail→pass | 18,344 | 25,344 | +38% | 1 | 1 | 0% | 2,100 | 1,815 | -14% | 0 | 0 | — |
case-12 | fail→pass | 34,536 | 27,018 | -22% | 1 | 1 | 0% | 1,374 | 1,014 | -26% | 0 | 0 | — |
case-13 | fail→pass | 16,153 | 21,307 | +32% | 1 | 1 | 0% | 1,689 | 1,390 | -18% | 0 | 0 | — |
case-14 | fail→pass | 42,309 | 11,113 | -74% | 1 | 1 | 0% | 1,159 | 1,630 | +41% | 0 | 0 | — |
case-15 | pass→pass | 19,971 | 7,922 | -60% | 1 | 1 | 0% | 2,332 | 1,755 | -25% | 0 | 0 | — |
case-16 | pass→pass | 21,115 | 2,851 | -86% | 1 | 1 | 0% | 1,750 | 1,198 | -32% | 0 | 0 | — |
case-17 | fail→pass | 16,132 | 8,200 | -49% | 1 | 1 | 0% | 1,727 | 1,239 | -28% | 0 | 0 | — |
case-18 | pass→pass | 18,910 | 7,381 | -61% | 1 | 1 | 0% | 2,199 | 1,133 | -48% | 0 | 0 | — |
case-19 | fail→fail | 20,667 | 10,480 | -49% | 1 | 1 | 0% | 2,358 | 1,920 | -19% | 0 | 0 | — |
case-20 | fail→pass | 21,881 | 17,884 | -18% | 1 | 1 | 0% | 3,496 | 3,129 | -10% | 0 | 0 | — |
case-21 | fail→fail | 34,420 | 27,071 | -21% | 1 | 1 | 0% | 8,217 | 6,381 | -22% | 0 | 0 | — |
case-22 | fail→fail | 23,388 | 23,597 | +1% | 1 | 1 | 0% | 4,409 | 4,019 | -9% | 0 | 0 | — |
case-23 | fail→fail | 17,409 | 15,645 | -10% | 1 | 1 | 0% | 3,409 | 2,764 | -19% | 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 22 counted toward the lift figure. The other 1 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 +48 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.