// Transform skill drafts into clear methodology through phased refinement that applies only what each skill needs. Use when: (1) after generating an initial skill draft, (2) reviewing a skill that reads as items rather than thinking, (3) before finalizing a skill for inclusion in a plugin, (4) after feedback that a skill lacks clarity.
| name | refine-skill-expression |
| description | Transform skill drafts into clear methodology through phased refinement that applies only what each skill needs. Use when: (1) after generating an initial skill draft, (2) reviewing a skill that reads as items rather than thinking, (3) before finalizing a skill for inclusion in a plugin, (4) after feedback that a skill lacks clarity. |
Transform skill drafts into clear methodology through phased refinement.
Phase 1: Diagnose
Invoke Skill("refine-verify") to assess the skill's current state. Verification identifies which refinement operations would meaningfully improve clarity, or concludes the skill is already clear enough.
If verification concludes refinement is complete, stop here.
Phase 2: Synthesize
If verification identifies question-lists, disconnected fragments, or checklist feel:
Skill("refine-synthesize") to transform these into flowing methodologySkill("refine-verify") to reassessAddress synthesis first when multiple issues exist—it often resolves other issues as a side effect.
Phase 3: Validate
Based on verification routing, invoke the appropriate subskill:
| Issue | Invoke |
|---|---|
| Domain-specific vocabulary | Skill("refine-test-transfer") |
| Introspection-based triggers | Skill("refine-validate-triggers") |
| Over-qualified triggers | Skill("refine-validate-triggers") |
| Negation-heavy instructions | Skill("refine-convert-negations") |
After each subskill completes, invoke Skill("refine-verify") to reassess before proceeding.
Phase 4: Conclude
When verification concludes the skill is clear enough, accept that conclusion. A skill with minor imperfections that reads clearly is better than one polished until its core insight is buried.
The goal is a skill that reads as someone explaining their thinking process—where each sentence flows into the next and the core methodology is apparent, not buried under process steps.
Scenario: A skill draft has question-lists and domain-specific vocabulary.
Flow:
Skill("refine-verify") → routes to synthesis (question-lists detected)Skill("refine-synthesize") → transforms questions into methodologySkill("refine-verify") → routes to test-transfer (domain vocabulary detected)Skill("refine-test-transfer") → revises vocabulary for cross-domain useSkill("refine-verify") → concludes skill is clear enoughScenario: A skill draft is already well-written with minor imperfections.
Flow:
Skill("refine-verify") → concludes further refinement risks overworkNo subskills invoked. The skill was already clear enough.
Scenario: A skill has question-lists and some negation-heavy phrasing.
Flow:
Skill("refine-verify") → routes to synthesisSkill("refine-synthesize") → transforms questions, incidentally converts negations to positive framingSkill("refine-verify") → concludes skill is clear enough (negation issue resolved)Synthesis addressed both issues, so no explicit negation pass was needed.