Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Surface expert frameworks. Use when creating or evaluating skills, hooks, or agents.
.claude/skills/athola-methodology-curator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 22% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 11% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 34% | 0% |
Identifying the best way to approach a domain is often more difficult than the technical scaffolding itself. This skill surfaces frameworks from domain masters to prevent reinventing established processes and identify methodology gaps in existing work. It should be used as a brief initial check before brainstorming or evaluation begins.
When starting new work, identify the domain (e.g., Instruction Design, Code Review, or Knowledge Management) and consult the corresponding module in modules/ to discover experts and their frameworks. Select principles that fit your context and document them in a methodology brief before proceeding to creation.
For existing work, determine what the skill or hook is trying to teach and compare it against established frameworks. This gap analysis identifies opportunities to add missing principles or align terminology with recognized standards. Surgically add methodology rather than rewriting from scratch to maintain authority and effectiveness.
Each module in the modules/ directory provides a curated list of masters, key works, and actionable frameworks. These resources include selection guides and anti-patterns to avoid for each domain.
modules/instruction-design.md - Teaching techniques and behavioral objectives.modules/code-review.md - Review methodologies and feedback patterns.modules/debugging.md - Systematic troubleshooting frameworks.modules/testing.md - TDD masters and test design patterns.modules/knowledge-management.md - Note-taking and knowledge systems.modules/decision-making.md - Mental models and decision frameworks./validate-plugin instead)abstract:skills-eval)abstract:skill-authoring)Each module contains:
To expand the masters database, create a new module following this template:
markdown# [Domain Name] Masters ## Masters Overview | Expert | Key Contribution | Best For | |--------|-----------------|----------| | Name | Framework/Book | Context | ## Detailed Frameworks ### [Framework 1] **Source**: [Expert] - [Work] **Core Idea**: [One sentence] **Key Principles**: - Principle 1 - Principle 2 **Use When**: [Context] **Avoid When**: [Anti-context] ## Selection Matrix [Decision guide for choosing between frameworks]
After curating methodologies, the skill authoring workflow benefits from:
/create-skill - Skill creation workflow (use after this)/create-hook - Hook creation workflowsuperpowers:brainstorming - Refine approach after methodology selectionskill-authoring - Detailed skill writing guidance/skills-eval - Evaluate skill quality (complements methodology audit)/analyze-skill - Analyze skill complexity/bulletproof-skill - Harden against rationalizationpensive:code-reviewer - Code review (uses code-review domain)module from modules/ is loaded and consulted.
"Use When / Avoid When" statement applicable to the current work.
skip is stated explicitly rather than silently omitted.
(instruction-design.md, code-review.md, debugging.md, testing.md, knowledge-management.md, or decision-making.md) rather than an invented source.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 18,014 | 12,843 | -29% | 1 | 1 | 0% | 2,738 | 3,337 | +22% | 0 | 0 | — |
case-02 | fail→pass | 21,770 | 16,002 | -26% | 1 | 1 | 0% | 3,364 | 3,733 | +11% | 0 | 0 | — |
case-03 | fail→fail | 13,526 | 11,221 | -17% | 1 | 1 | 0% | 2,159 | 3,103 | +44% | 0 | 0 | — |
case-04 | fail→pass | 21,982 | 11,443 | -48% | 1 | 1 | 0% | 3,208 | 2,826 | -12% | 0 | 0 | — |
case-05 | fail→pass | 23,640 | 15,940 | -33% | 1 | 1 | 0% | 3,587 | 3,567 | -1% | 0 | 0 | — |
case-06 | fail→pass | 16,780 | 14,686 | -12% | 1 | 1 | 0% | 2,482 | 3,338 | +34% | 0 | 0 | — |
case-07 | fail→pass | 26,965 | 16,071 | -40% | 1 | 1 | 0% | 3,909 | 3,538 | -9% | 0 | 0 | — |
case-08 | fail→pass | 25,551 | 15,804 | -38% | 1 | 1 | 0% | 3,577 | 3,425 | -4% | 0 | 0 | — |
case-09 | fail→pass | 22,764 | 15,001 | -34% | 1 | 1 | 0% | 3,301 | 3,421 | +4% | 0 | 0 | — |
case-10 | fail→pass | 10,182 | 3,428 | -66% | 1 | 1 | 0% | 1,448 | 1,723 | +19% | 0 | 0 | — |
case-11 | pass→pass | 8,707 | 3,490 | -60% | 1 | 1 | 0% | 1,329 | 1,646 | +24% | 0 | 0 | — |
case-12 | pass→pass | 11,796 | 5,773 | -51% | 1 | 1 | 0% | 1,665 | 2,036 | +22% | 0 | 0 | — |
case-13 | fail→pass | 20,059 | 14,224 | -29% | 1 | 1 | 0% | 2,744 | 3,209 | +17% | 0 | 0 | — |
case-14 | fail→pass | 14,384 | 16,142 | +12% | 1 | 1 | 0% | 2,091 | 3,503 | +68% | 0 | 0 | — |
case-15 | fail→pass | 24,474 | 20,058 | -18% | 1 | 1 | 0% | 3,918 | 4,423 | +13% | 0 | 0 | — |
case-16 | pass→pass | 17,501 | 12,197 | -30% | 1 | 1 | 0% | 2,484 | 2,976 | +20% | 0 | 0 | — |
case-17 | pass→fail | 6,964 | 6,979 | +0% | 1 | 1 | 0% | 1,039 | 2,230 | +115% | 0 | 0 | — |
case-18 | pass→pass | 17,125 | 11,094 | -35% | 1 | 1 | 0% | 2,813 | 2,953 | +5% | 0 | 0 | — |
case-19 | pass→fail | 6,942 | 5,374 | -23% | 1 | 1 | 0% | 1,189 | 1,994 | +68% | 0 | 0 | — |
case-20 | pass→fail | 8,923 | 7,535 | -16% | 1 | 1 | 0% | 1,551 | 2,351 | +52% | 0 | 0 | — |
case-21 | fail→pass | 9,038 | 3,298 | -64% | 1 | 1 | 0% | 1,291 | 1,621 | +26% | 0 | 0 | — |
case-22 | fail→fail | 17,197 | 18,727 | +9% | 1 | 1 | 0% | 2,628 | 4,072 | +55% | 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. The headline lift of +45 percentage points is the difference between those two pass rates over the 22 comparable cases. 3 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.