Skip to content

스킬 라이프사이클 관리 — skill_nature 태깅 + 경량화 체계 #145

@bluejayA

Description

@bluejayA

배경

모델 발전에 따라 보상(compensation) 스킬의 활용 가치가 감소하는 시점에 대비하여, 스킬 라이프사이클 관리 체계를 도입한다.

설계 문서

  • docs/research/2026-04-06-skill-lifecycle-strategy.md

범위 (Phase 1 MVP)

  1. devflow-conventions.mdskill_nature, lifecycle, model_dependency 규약 추가
  2. 28개 SKILL.md에 태그 일괄 추가 (compensation/amplification/hybrid 분류)
  3. superpowers-tracking에 Compensation Decay 분석 템플릿 추가
  4. validate-skills.sh에 새 필드 검증 (warning 레벨)

Phase 2 (후속)

  1. 첫 번째 lightening 후보 실행: verification-before-completion
  2. audit 기반 skip-rate 분석 자동화

핵심 개념

  • compensation: 모델 약점 보상 → 모델 발전 시 경량화 후보
  • amplification: 조직/개인 고유 지식 → 모델과 무관하게 영구 유지
  • lifecycle: draft → active → lightened → absorbed | archived

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions