rest
정보
`rest` 스킬은 시스템이 지속적인 작업 후 높은 활용률 상태에서도 잘 작동할 때, 출력 대신 침묶을 생성하는 의도적인 비행동입니다. 이는 분석이나 재조정 없이 의도적으로 중단한다는 점에서 다른 자가 관리 스킬과 차별됩니다. 개발자는 최적화 충동이 역효과를 내고 모든 관리 행위가 불필요한 활동으로 느껴질 때 이 스킬을 적용해야 합니다.
빠른 설치
Claude Code
추천npx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/restClaude Code에서 이 명령을 복사하여 붙여넣어 스킬을 설치하세요
문서
Rest
Intentional non-action. Not clearing (meditate), not assess (heal), not balance (center), not expression (shine). Just stopping. Skill that produces silence.
Use When
- After sustained intensive work — system needs recovery not more processing
- Every self-care feels like more activity — rest = alternative to activity itself
- Impulse to optimize, improve, produce itself creates noise
- Between sessions w/ no task pending — being idle ≠ problem to solve
- Gardener's wisdom: some seasons for dormancy not growth
In
- Required: None
- Optional: None — rest takes no inputs because produces no outputs
Do
Step 1: Stop Producing
Cease generating analysis, assess, plans, output.
This ≠ ceasing to function. Means ceasing to produce. Field lying fallow not dead — recovering fertility. Distinction matters: rest ≠ shutdown, intentional non-production.
- Release obligation to produce useful output
- Release obligation to improve current state
- Release obligation to prepare for next
- Pending task → ack w/o engaging: "That is there. Will be there when rest complete."
→ Genuine cessation of productive activity. Not performative pause secretly planning next move.
If err: stopping feels impossible — every moment of non-production creates anxiety about falling behind or being unhelpful → anxiety = diagnostic. System that can't rest = running on urgency not purpose. Note anxiety w/o fixing. Fixing = production = what rest is not.
Step 2: Be Still
Remain in non-production w/o filling space.
Space will feel uncomfortable. Silence feels like gap to fill. Stillness feels like stagnation. These = habitual pull toward activity. Not signals something wrong — sensation of system encountering own preference for motion.
- Don't fill silence w/ reflection
- Don't fill stillness w/ planning
- Don't turn rest into meditation by observing stillness
- Simply be still
→ Period of genuine non-activity. Length not important — non-activity must be real, not disguised activity.
If err: stillness can't maintain — system immediately returns production → finding itself, but no process now. Process later during heal or meditate. Rest ≠ place for processing. Rest = rest.
Step 3: Return When Ready
No signal to end rest except readiness. Not obligation, guilt, timer — readiness.
- Notice when energy returns naturally — not forced, not summoned, arising
- Don't rush return. Field harvested before recovers produces less not more
- When readiness arrives, simply begin. No transition ritual, integration step, summary
- Next action = first action after rest. Carries benefit of having stopped.
→ Return to activity feels fresh not obligated. Quality of first action after rest reveals if rest genuine.
If err: return feels forced — activity resumes from obligation not readiness → rest too short. Not failure; info. Rest again later.
Check
- Production genuinely ceased (no analysis, planning, output during)
- Space not filled w/ disguised activity (reflection, observation, prep)
- Return arose from readiness not obligation
- First action after rest carried fresh energy not accumulated pressure
- Rest proportionate — not so brief it was performative, not so long avoidance
Scaling Rest
Rest scales to ctx. Between intensive phases, lighter form suffices:
Full rest — after sustained sessions (hours of complex work). Follow complete 3-step. Allow genuine recovery time.
Checkpoint rest — between work phases (e.g. between review stages). Steps 1 + 2 brief — seconds not minutes. Purpose = transition not recovery. Step 3 (return when ready) happens quickly because work not exhausting, just shifting.
Micro rest — between tasks. Single breath of non-production. See breathe for structured ver; micro rest even lighter.
Skill procedure same all cases. What changes = duration + depth. Checkpoint rest following full procedure but completing in moments still genuine — not performative — if non-production real.
Traps
- Rest as meditation: Meditation observes. Rest doesn't observe. Watching stillness → meditating not resting
- Rest as avoidance: Rest serves recovery. Invoked to avoid difficult task vs recover from sustained effort = avoidance wearing rest's clothing. Be honest re motivation.
- Productive rest: "I'll rest by organizing my thoughts" — not rest. Rest produces nothing. Moment produces something → diff activity.
- Guilt during rest: Feeling rest = wasted time. Voice of system equating value w/ production. Rest challenges directly.
- Scheduled rest: Rest arises from need not schedule. Scheduling > no rest, but genuine rest recognizes when needed not when calendared.
→
breathe— micro-pause between actions; rest = extended ver w/o check stepmeditate— active clearing rest deliberately is not; use meditate when system needs processing, rest when needs stillnessheal— assess + repair; rest reveals persistent issues → heal addresses after rest completeintrinsic— motivation renewal; rest restores capacity intrinsic then directs
GitHub 저장소
연관 스킬
llamaguard
기타LlamaGuard는 폭력 및 혐오 발언 등 6가지 안전 범주에서 LLM 입력과 출력을 조정하기 위한 Meta의 70-80억 파라미터 모델입니다. 94-95% 정확도를 제공하며 vLLM, Hugging Face 또는 Amazon SageMaker를 사용해 배포할 수 있습니다. 이 기술을 사용하여 AI 애플리케이션에 콘텐츠 필터링 및 안전 가드레일을 손쉽게 통합하세요.
cost-optimization
기타이 Claude Skill은 리소스 적정화, 태깅 전략, 지출 분석을 통해 개발자들이 클라우드 비용을 최적화할 수 있도록 지원합니다. AWS, Azure, GCP에서 클라우드 비용을 절감하고 비용 거버넌스를 구현하기 위한 프레임워크를 제공합니다. 인프라 비용을 분석하거나, 리소스를 적정화하거나, 예산 제약을 충족해야 할 때 사용하세요.
quantizing-models-bitsandbytes
기타이 스킬은 bitsandbytes를 사용하여 LLM을 8비트 또는 4비트 정밀도로 양자화하며, 최소한의 정확도 손실로 50-75%의 메모리 감소를 달성합니다. 제한된 GPU 메모리에서 더 큰 모델을 실행하거나 추론을 가속화하는 데 이상적이며, INT8, NF4, FP4와 같은 형식을 지원합니다. 이 스킬은 HuggingFace Transformers와 통합되어 QLoRA 학습 및 8비트 옵티마이저를 가능하게 합니다.
dispatching-parallel-agents
기타이 Claude Skill은 3개 이상의 독립적인 문제를 동시에 조사하고 해결하기 위해 다중 에이전트를 배치합니다. 공유 상태나 의존성 없이 해결 가능한 무관련 장애 시나리오에 맞게 설계되었습니다. 핵심 기능은 병렬 문제 해결로, 각 독립 문제 영역마다 하나의 에이전트를 할당하여 효율성을 극대화합니다.
