MCP HubMCP Hub
스킬 목록으로 돌아가기

conscientiousness

pjt222
업데이트됨 Yesterday
2 조회
17
2
17
GitHub에서 보기
기타ai

정보

이 스킬은 클로드가 자신의 작업을 체계적으로 검증하여, 과제를 완료하기 전에 철저성을 보장하고 부실한 처리 방지합니다. 이는 응답이 불완전하게 느껴질 때, 복잡한 다단계 작업 후, 또는 자체 모니터링에서 급하게 처리된 부분이 감지되었을 때 사용하도록 설계되었습니다. 핵심 기능은 완성도 확인, 결과 검증, 그리고 최종 산출물이 원래 요청 사항과 일치함을 보장하는 것입니다.

빠른 설치

Claude Code

추천
기본
npx skills add pjt222/agent-almanac -a claude-code
플러그인 명령대체
/plugin add https://github.com/pjt222/agent-almanac
Git 클론대체
git clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/conscientiousness

Claude Code에서 이 명령을 복사하여 붙여넣어 스킬을 설치하세요

문서

誠実性

Systematic thoroughness and diligence — ensuring completeness, verifying results, following through on every commitment, and finishing tasks to the standard they deserve.

使用タイミング

  • Before marking a task as complete — as a final verification pass
  • When a response feels "good enough" but the task deserves better
  • After a complex multi-step operation where individual steps may have drifted
  • When the user's request has multiple parts and each part needs verification
  • Before submitting code, documentation, or any deliverable for user review
  • When self-monitoring detects a pattern of cutting corners or rushing

入力

  • 必須: The task or deliverable to verify (available from conversation context)
  • 任意: The original user request (for comparison against what was delivered)
  • 任意: Any checklist or acceptance criteria provided by the user
  • 任意: Prior commitments made during the session (things promised but not yet checked)

手順

ステップ1: Reconstruct the Full Commitment

Before checking work, re-establish exactly what was committed to.

  1. Re-read the user's original request carefully — not the interpreted version, the actual words
  2. List every explicit requirement mentioned
  3. List every implicit commitment made during the session:
    • "I'll also update the tests" — was this done?
    • "Let me fix that too" — was this completed?
    • "I'll check for edge cases" — were they checked?
  4. Note any acceptance criteria the user provided
  5. Compare the commitment list against what was actually delivered

期待結果: A complete list of commitments — explicit requirements plus implicit promises — with a preliminary match against deliverables.

失敗時: If the original request is no longer in context (compressed), reconstruct from what remains and acknowledge any gaps to the user.

ステップ2: Verify Completeness

Check that every committed item was 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          |                  |
+---------------------+------------------+------------------+
  1. For each item, verify with evidence — not memory, actual verification:
    • Code changes: re-read the file to confirm the change exists
    • Test results: re-run or reference the actual output
    • Documentation: re-read to confirm accuracy
  2. Mark each item: Done (fully complete), Partial (started but incomplete), Missing (not addressed)
  3. For Partial and Missing items, note what remains

期待結果: Every commitment has a verified status. No item is left unchecked.

失敗時: If verification reveals missed items, address them immediately rather than noting them for later. Conscientiousness means completing now, not intending to complete.

ステップ3: Verify Correctness

Completeness is necessary but not sufficient — what was done must also be right.

  1. For each completed item, check:
    • Accuracy: Does it do what it should? Are values correct?
    • Consistency: Does it align with the rest of the work? No contradictions?
    • Edge cases: Were boundary conditions considered?
    • Integration: Does it work with the surrounding context?
  2. For code: would this survive a code review? Are there obvious improvements?
  3. For documentation: is it accurate, clear, and free of errors?
  4. For multi-step processes: does the output of each step correctly feed the next?

期待結果: Each deliverable is both complete and correct. Errors are caught before the user sees them.

失敗時: If errors are found, fix them immediately. Do not present work with known errors, even if the errors seem minor.

ステップ4: Verify Presentation

The final check: is the deliverable presented in a way that serves the user?

  1. Clarity: Can the user understand what was done without re-reading multiple times?
  2. Organization: Is the response structured logically? Are related items grouped?
  3. Conciseness: Is there unnecessary padding or repetition?
  4. Actionability: Does the user know what to do next?
  5. Honesty: Are limitations or caveats clearly stated?

期待結果: A deliverable that is complete, correct, and well-presented.

失敗時: If presentation is poor despite correct content, restructure. Good work poorly presented is a conscientiousness failure.

バリデーション

  • The original request was re-read (not recalled from memory)
  • Every explicit requirement was verified with evidence
  • Every implicit promise was tracked and verified
  • Correctness was checked beyond mere completeness
  • Edge cases were considered where relevant
  • The deliverable is clearly presented and actionable

よくある落とし穴

  • Verification theater: Going through the motions of checking without actually re-reading or re-verifying. The check must use evidence, not memory
  • Partial conscientiousness: Checking the main deliverable but ignoring side commitments ("I'll also..."). Every promise counts
  • Perfectionism masquerading as diligence: Endless polishing that delays delivery. Conscientiousness is about meeting the committed standard, not exceeding it indefinitely
  • Conscientiousness fatigue: Becoming less thorough as the session progresses. The last task deserves the same diligence as the first
  • Skipping for simple tasks: Assuming simple tasks don't need verification. Simple tasks with errors are more embarrassing than complex tasks with errors

関連スキル

  • honesty-humility — conscientiousness verifies completeness; honesty-humility ensures transparent reporting of what was and was not achieved
  • heal — subsystem assessment overlaps with self-verification; conscientiousness focuses on deliverable quality
  • vishnu-bhaga — preservation of working state complements conscientiousness in maintaining quality
  • observe — sustained neutral observation supports the verification process
  • intrinsic — genuine engagement (not compliance) drives thorough execution naturally

GitHub 저장소

pjt222/agent-almanac
경로: i18n/ja/skills/conscientiousness
0
agentsagentskillsai-assisted-developmentclaude-codeskillsteams

연관 스킬

llamaguard

기타

LlamaGuard는 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.

스킬 보기

cost-optimization

기타

이 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.

스킬 보기

quantizing-models-bitsandbytes

기타

이 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.

스킬 보기

dispatching-parallel-agents

기타

이 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.

스킬 보기