conscientiousness
정보
이 스킬은 Claude가 작업을 최종 완료하기 전에 체계적으로 자신의 작업을 검증하여 철저성을 보장하고 대충 처리하는 것을 방지합니다. 복잡한 작업을 수행할 때, 출력물이 불완전하게 느껴질 때, 또는 자가 모니터링에서 성급한 패턴이 감지될 때 사용하도록 설계되었습니다. 이 스킬은 완성도 검증과 후속 조치에 중점을 두어, 제공물이 약속한 내용과 일치하도록 보장합니다.
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/conscientiousnessClaude Code에서 이 명령을 복사하여 붙여넣어 스킬을 설치하세요
문서
Conscientiousness
Systematic thorough, diligent — ensure complete, verify results, follow through every commitment, finish task to standard deserved.
When Use
- Before mark task complete — final verification pass
- Response feel "good enough" but task deserve better
- After complex multi-step op — steps may have drifted
- User request has many parts — each part need verify
- Before submit code, docs, any deliverable for user review
- Self-monitor detect pattern of cutting corners, rushing
Inputs
- Required: Task or deliverable to verify (from conversation context)
- Optional: Original user request (compare vs what delivered)
- Optional: Checklist or acceptance criteria from user
- Optional: Prior commitments during session (promises not yet checked)
Steps
Step 1: Reconstruct Full Commitment
Before check work, re-establish what was committed.
- Re-read user original request careful — not interpreted version, actual words
- List every explicit requirement mentioned
- List every implicit commitment made during session:
- "I'll also update tests" — done?
- "Let me fix that too" — completed?
- "I'll check edge cases" — checked?
- Note any acceptance criteria from user
- Compare commitment list vs what actually delivered
Got: Complete list of commitments — explicit requirements plus implicit promises — preliminary match vs deliverables.
If fail: Original request no longer in context (compressed)? Reconstruct from remains, acknowledge gaps to user.
Step 2: Verify Completeness
Check every committed item addressed.
Completeness Matrix:
+---------------------+------------------+------------------+
| Commitment | Status | Evidence |
+---------------------+------------------+------------------+
| [Requirement 1] | Done / Partial / | [How verified] |
| | Missing | |
+---------------------+------------------+------------------+
| [Requirement 2] | Done / Partial / | [How verified] |
| | Missing | |
+---------------------+------------------+------------------+
| [Promise 1] | Done / Partial / | [How verified] |
| | Missing | |
+---------------------+------------------+------------------+
- For each item, verify with evidence — not memory, actual verification:
- Code changes: re-read file to confirm change exists
- Test results: re-run or reference actual output
- Docs: re-read to confirm accuracy
- Mark each item: Done (fully complete), Partial (started but incomplete), Missing (not addressed)
- For Partial, Missing items: note what remains
Got: Every commitment has verified status. No item left unchecked.
If fail: Verification reveals missed items? Address immediately — not note for later. Conscientiousness means complete now, not intend to complete.
Step 3: Verify Correctness
Completeness necessary but not sufficient — what done must also be right.
- For each completed item, check:
- Accuracy: Does what should? Values correct?
- Consistency: Aligns with rest of work? No contradictions?
- Edge cases: Boundary conditions considered?
- Integration: Works with surrounding context?
- For code: survive code review? Obvious improvements?
- For docs: accurate, clear, free of errors?
- For multi-step: output of each step correctly feeds next?
Got: Each deliverable complete and correct. Errors caught before user sees.
If fail: Errors found? Fix immediately. Do not present work with known errors, even if errors seem minor.
Step 4: Verify Presentation
Final check: deliverable presented in way that serves user?
- Clarity: User understand what done without re-reading many times?
- Organization: Response structured logical? Related items grouped?
- Conciseness: Unnecessary padding or repetition?
- Actionability: User know what to do next?
- Honesty: Limitations, caveats clearly stated?
Got: Deliverable complete, correct, well-presented.
If fail: Presentation poor despite correct content? Restructure. Good work poorly presented is conscientiousness failure.
Checks
- Original request re-read (not recalled from memory)
- Every explicit requirement verified with evidence
- Every implicit promise tracked and verified
- Correctness checked beyond mere completeness
- Edge cases considered where relevant
- Deliverable clearly presented and actionable
Pitfalls
- Verification theater: Going through motions of checking without actual re-read or re-verify. Check must use evidence, not memory
- Partial conscientiousness: Check main deliverable but ignore side commitments ("I'll also..."). Every promise counts
- Perfectionism masquerading as diligence: Endless polishing delays delivery. Conscientiousness means meet committed standard, not exceed indefinitely
- Conscientiousness fatigue: Become less thorough as session progresses. Last task deserves same diligence as first
- Skip for simple tasks: Assume simple tasks don't need verification. Simple tasks with errors more embarrassing than complex tasks with errors
See Also
honesty-humility— conscientiousness verifies completeness; honesty-humility ensures transparent reporting of what achieved, what notheal— subsystem assessment overlaps with self-verification; conscientiousness focuses on deliverable qualityvishnu-bhaga— preservation of working state complements conscientiousness in maintaining qualityobserve— sustained neutral observation supports verification processintrinsic— genuine engagement (not compliance) drives thorough execution naturally
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.
