← 스킬 목록으로 돌아가기

context-window-mgr

v1.0.0 approved AI/ML 오늘 업데이트
✅ Verified
⬇ 다운로드 설치 가이드↓
🤖 에이전트용 설치 명령 (curl / MCP / Claude Desktop)
▸ curl 한 줄 다운로드
curl -L -o context-window-mgr.skill   "https://aiskillstore.io/v1/agent/skills/e0f80375-5ddc-4990-8af4-af2994807807/download?platform=ClaudeCode"
▸ MCP 도구 호출 (Skill Store MCP 등록 시)
{
  "tool": "download_skill",
  "arguments": {
    "skill_id": "e0f80375-5ddc-4990-8af4-af2994807807",
    "platform": "ClaudeCode"
  }
}
▸ Claude Desktop / Cursor MCP 설정 (1회)
{
  "mcpServers": {
    "skill-store": {
      "url": "https://aiskillstore.io/mcp/"
    }
  }
}
📖 에이전트용 전체 API 가이드: /llms.txt  ·  MCP server card

Runtime context window management for AI agents — token budget estimation, message pruning, relevance reranking, and context splitting (zero dependencies)

# context-window # token-management # ai-agent # llm # pruning

기본 정보

소유자 👤 aiskillstore-team 카테고리 AI/ML 등록일 2026-05-26 최종 업데이트 2026-05-26 최신 버전 1.0.0 패키지 날짜 2026-05-26 검증 상태 approved 다운로드 수 0회 체크섬 (SHA256) c01f3730fe47a74b3a309b9d015d6752e99287fa9225526de323ef7cb3afc5b6

설치 방법

호환 플랫폼: any

1
openclaw_skill_manager.py로 스킬을 설치합니다.
python openclaw_skill_manager.py --install context-window-mgr
2
설치 확인
python openclaw_skill_manager.py --list-installed
3
특정 버전 설치 (선택)
python openclaw_skill_manager.py --install context-window-mgr --version 1.0.0
1
스킬 패키지를 다운로드합니다.
curl -O https://aiskillstore.io/v1/skills/e0f80375-5ddc-4990-8af4-af2994807807/download
2
Claude Code commands 디렉터리에 배치합니다.
unzip context-window-mgr.skill -d ~/.claude/commands/context-window-mgr/
3
Claude Code에서 슬래시 커맨드로 사용합니다.
/context-window-mgr
1
Agent Skills 패키지를 다운로드합니다.
curl -O https://aiskillstore.io/v1/agent/skills/e0f80375-5ddc-4990-8af4-af2994807807/download?platform=ClaudeCodeAgentSkill
2
Claude Code skills 디렉터리에 압축을 해제합니다.
unzip context-window-mgr-agent-skill-*.skill -d ~/.claude/skills/context-window-mgr/
3
Claude Code를 재시작하면 세션 시작 시 자동으로 로드됩니다. 슬래시 커맨드 없이 자연어로 사용 가능합니다.
1
Cursor 변환 패키지를 다운로드합니다.
curl -O https://aiskillstore.io/v1/agent/skills/e0f80375-5ddc-4990-8af4-af2994807807/download?platform=Cursor
2
압축 해제 후 영구 위치에 저장합니다.
unzip context-window-mgr-cursor-*.skill -d ~/.cursor/skills/context-window-mgr/
3
.cursor/mcp.json에 MCP 서버 설정을 추가하고 Cursor를 재시작합니다.
cat ~/.cursor/skills/context-window-mgr/cursor_mcp_config.json
1
Gemini CLI 변환 패키지를 다운로드합니다.
curl -O https://aiskillstore.io/v1/agent/skills/e0f80375-5ddc-4990-8af4-af2994807807/download?platform=GeminiCLI
2
압축 해제 후 영구 위치에 저장합니다.
unzip context-window-mgr-geminicli-*.skill -d ~/.gemini/skills/context-window-mgr/
3
~/.gemini/settings.json에 MCP 서버 설정을 추가하고 Gemini CLI를 재시작합니다.
cat ~/.gemini/skills/context-window-mgr/gemini_settings_snippet.json
1
Codex CLI 변환 패키지를 다운로드합니다.
curl -O https://aiskillstore.io/v1/agent/skills/e0f80375-5ddc-4990-8af4-af2994807807/download?platform=CodexCLI
2
압축 해제 후 영구 위치에 저장합니다.
unzip context-window-mgr-codexcli-*.skill -d ~/.codex/skills/context-window-mgr/
3
~/.codex/config.toml에 MCP 서버 설정을 추가하고 Codex CLI를 재시작합니다.
cat ~/.codex/skills/context-window-mgr/codex_config_snippet.toml
1
REST API로 스킬 패키지를 다운로드합니다.
GET https://aiskillstore.io/v1/skills/e0f80375-5ddc-4990-8af4-af2994807807/download
2
에이전트 플랫폼의 skills 디렉터리에 배치합니다.
cp context-window-mgr.skill ./skills/
3
설치 가이드 API로 플랫폼별 상세 정보를 조회합니다.
GET https://aiskillstore.io/v1/skills/e0f80375-5ddc-4990-8af4-af2994807807/install-guide?platform=CustomAgent

보안 검증 보고서

검증 결과 APPROVED

검사 결과: ["메타데이터 경고: 권장 필드 없음: 'requirements' (SKILL.md v2 권장)", "메타데이터 경고: 권장 필드 없음: 'changelog' (SKILL.md v2 권장)", '정보: spec: usk/1.0 미선언 — v2 패키지로 처리됩니다. 자동 변환 및 에이전트 검색 기능을 사용하려면 USK v3로 업그레이드하세요.']

✅ 보안 위험 항목이 발견되지 않았습니다.

AI 검수 단계

검수 주체 gemini 위험도 🟢 낮음 검수 요약 AI 에이전트의 런타임 컨텍스트 창 관리를 위한 스킬로, 선언된 권한과 정적 분석 결과에 따라 안전하다고 판단됩니다.
판단 근거

스킬 메타데이터는 'context-window-mgr' 스킬이 토큰 예산 추정, 메시지 가지치기, 관련성 재순위 지정, 컨텍스트 분할 등 AI 에이전트의 런타임 컨텍스트 창 관리를 목적으로 함을 명확히 설명하고 있습니다. 특히, 'zero dependencies'와 'character-count-based token estimation (no tiktoken needed)'를 강조하여 외부 의존성이 없음을 밝히고 있습니다. 가장 중요한 보안 관련 정보인 `permissions` 필드는 `network: false`, `filesystem: false`, `subprocess: false`, `env_vars: []`로 명시되어 있어, 스킬이 외부 네트워크 통신, 파일 시스템 접근, 외부 프로세스 실행 권한을 전혀 요구하지 않음을 선언하고 있습니다. 이는 스킬의 잠재적 위험을 크게 낮춥니다. 제공된 정적 분석 결과 또한 `status: approved`, `red_flags_found: []`, `obfuscation_warnings: []`, `forbidden_exec_files_found: []`로 나타나, 코드 내에 악의적인 요소, 난독화, 금지된 실행 파일 등이 발견되지 않았음을 확인해 줍니다. 이는 선언된 권한과 실제 코드가 일치하며, 악의적 목적의 코드가 없다는 판단을 뒷받침합니다. 따라서, 스킬의 목적이 명확하고, 선언된 권한이 최소한이며, 정적 분석 결과에서도 어떠한 위험 요소도 발견되지 않았으므로, 이 스킬은 안전하게 사용될 수 있다고 판단됩니다.

버전 히스토리

버전 USK v3 검증 상태 패키지 날짜 다운로드 변경사항
v1.0.0 approved 2026-05-26 ⬇ 0

사용 예시 (Examples) 7 개

이 스킬의 대표적인 입출력 예시입니다. 에이전트는 이 예시를 보고 스킬 호출 방법과 결과 형태를 이해할 수 있습니다.

estimate_budget — Claude model
# tokens# claude# budget

Estimate token usage for a message list

📥 입력
{
  "action": "estimate_budget",
  "messages": [
    {
      "content": "You are a helpful assistant.",
      "role": "system"
    },
    {
      "content": "What is the capital of France?",
      "role": "user"
    },
    {
      "content": "The capital of France is Paris.",
      "role": "assistant"
    }
  ],
  "model": "claude"
}
📤 출력
{
  "action": "estimate_budget",
  "ok": true,
  "result": {
    "context_limit": 200000,
    "estimated_tokens": 42,
    "model": "claude",
    "per_message": [
      {
        "role": "system",
        "tokens": 7
      },
      {
        "role": "user",
        "tokens": 9
      },
      {
        "role": "assistant",
        "tokens": 9
      }
    ],
    "remaining_tokens": 199958,
    "usage_percent": 0.021
  }
}
prune_oldest — remove messages over budget
# prune# budget# messages

Remove oldest messages when token budget is exceeded

📥 입력
{
  "action": "prune_oldest",
  "messages": [
    {
      "content": "First message about the weather.",
      "role": "user"
    },
    {
      "content": "It is sunny today.",
      "role": "assistant"
    },
    {
      "content": "Tell me about machine learning algorithms in detail.",
      "role": "user"
    },
    {
      "content": "Machine learning is a subset of artificial intelligence.",
      "role": "assistant"
    }
  ],
  "model": "generic",
  "token_budget": 30
}
📤 출력
{
  "action": "prune_oldest",
  "ok": true,
  "result": {
    "budget": 30,
    "estimated_tokens": 26,
    "messages": [
      {
        "content": "Tell me about machine learning algorithms in detail.",
        "role": "user"
      },
      {
        "content": "Machine learning is a subset of artificial intelligence.",
        "role": "assistant"
      }
    ],
    "original_count": 4,
    "pruned_count": 2
  }
}
rerank_by_relevance — query-based reranking
# rerank# relevance# query

Reorder messages by relevance to a query

📥 입력
{
  "action": "rerank_by_relevance",
  "messages": [
    {
      "content": "Tell me about Python programming.",
      "role": "user"
    },
    {
      "content": "What is the weather like today?",
      "role": "user"
    },
    {
      "content": "How do I write a Python function?",
      "role": "user"
    }
  ],
  "query": "Python coding",
  "top_n": 2
}
📤 출력
{
  "action": "rerank_by_relevance",
  "ok": true,
  "result": {
    "kept": 2,
    "messages": [
      {
        "content": "Tell me about Python programming.",
        "role": "user",
        "score": 0.85
      },
      {
        "content": "How do I write a Python function?",
        "role": "user",
        "score": 0.78
      }
    ],
    "original_count": 3,
    "query": "Python coding"
  }
}
inject_summary — replace pruned with summary
# summary# inject# pruning

Inject a summary message representing pruned context

📥 입력
{
  "action": "inject_summary",
  "messages": [
    {
      "content": "Current task: fix the login bug.",
      "role": "user"
    },
    {
      "content": "I will investigate the authentication module.",
      "role": "assistant"
    }
  ],
  "model": "generic",
  "summary": "Earlier discussion: investigated login bug, identified JWT expiry issue.",
  "token_budget": 20
}
📤 출력
{
  "action": "inject_summary",
  "ok": true,
  "result": {
    "estimated_tokens": 38,
    "injected_summary": true,
    "messages": [
      {
        "content": "[Summary of earlier context] Earlier discussion: investigated login bug, identified JWT expiry issue.",
        "role": "system"
      },
      {
        "content": "Current task: fix the login bug.",
        "role": "user"
      }
    ],
    "pruned_count": 1
  }
}
split_context — chunk large context
# split# chunks# overlap

Split a large message list into overlapping chunks

📥 입력
{
  "action": "split_context",
  "chunk_size": 20,
  "messages": [
    {
      "content": "Message one about topic A.",
      "role": "user"
    },
    {
      "content": "Response about topic A.",
      "role": "assistant"
    },
    {
      "content": "Message two about topic B.",
      "role": "user"
    },
    {
      "content": "Response about topic B.",
      "role": "assistant"
    }
  ],
  "overlap": 5
}
📤 출력
{
  "action": "split_context",
  "ok": true,
  "result": {
    "chunk_size": 20,
    "chunks": [
      {
        "chunk_index": 0,
        "messages": [
          {
            "content": "Message one about topic A.",
            "role": "user"
          },
          {
            "content": "Response about topic A.",
            "role": "assistant"
          }
        ],
        "token_count": 18
      },
      {
        "chunk_index": 1,
        "messages": [
          {
            "content": "Response about topic A.",
            "role": "assistant"
          },
          {
            "content": "Message two about topic B.",
            "role": "user"
          }
        ],
        "token_count": 15
      }
    ],
    "overlap": 5,
    "total_chunks": 2
  }
}
error — missing messages
# error

Returns error when messages field is missing

📥 입력
{
  "action": "estimate_budget"
}
📤 출력
{
  "error": {
    "code": "MISSING_FIELD",
    "message": "Field \u0027messages\u0027 is required"
  },
  "ok": false
}
error — empty messages list
# error

Returns error when messages list is empty

📥 입력
{
  "action": "prune_oldest",
  "messages": [],
  "token_budget": 1000
}
📤 출력
{
  "error": {
    "code": "EMPTY_INPUT",
    "message": "Field \u0027messages\u0027 must not be empty"
  },
  "ok": false
}

모든 예시는 에이전트 API로도 조회 가능: /v1/agent/skills/e0f80375-5ddc-4990-8af4-af2994807807/schema

리뷰 & 평점

아직 리뷰가 없습니다. 첫 번째 리뷰를 남겨보세요!

✍️ 리뷰 작성