Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/sonaiengine/gwanjong-mcp/claude-mdgit clone --depth 1 https://github.com/SonAIengine/gwanjong-mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/sonaiengine/gwanjong-mcp/claude-md)<a href="https://agentmods.dev/instructions/sonaiengine/gwanjong-mcp/claude-md"><img src="https://agentmods.dev/badge/instructions/sonaiengine/gwanjong-mcp/claude-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.03295 | $0.03295 |
| Opus 5 | $0.01648 | $0.01648 |
| Sonnet 5 | $0.00659 | $0.00659 |
| Haiku 4.5 | $0.00330 | $0.00330 |
Grade A, and why
gwanjong-mcp CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 277 lines — stays where its author put it; the contents beside it link to each section on GitHub.
gwanjong-mcp
AI Growth OS — Stateful Pipeline MCP 서버 + 자율 소셜 에이전트. 8개 tool + EventBus 플러그인 아키텍처.
핵심 설계: 토큰 효율 + 느슨한 결합
- Tool 8개 — description이 매 호출마다 시스템 프롬프트에 포함되므로 최소화
- 서버 상태 유지 — scout 결과를 GwanjongState에 캐시, ID 참조로 후속 호출
- 서버가 분석 담당 — trending/search/filter/score는 서버 내부 처리, LLM은 판단+생성만
- EventBus — 모듈 간 직접 import 없이 이벤트 pub/sub. 플러그인 유무와 무관하게 동작
- 캠페인 기반 — 모든 활동을 캠페인 단위로 추적, UTM 태깅, 전환 측정
프로젝트 구조
gwanjong_mcp/
├── __init__.py # 패키지
├── __main__.py # python -m gwanjong_mcp
├── server.py # PipelineMCP + 8 tools + GwanjongState + EventBus 조립
├── events.py # EventBus + Event + Blocked (유일한 공유 지점)
├── pipeline.py # scout/draft/strike 코어 (registry + EventBus)
├── types.py # Opportunity, DraftContext, ActionRecord, Campaign, Asset, ...
├── storage.py # SQLite 스키마 (actions, campaigns, conversions, assets, schedule, ...)
├── setup.py # 플랫폼 온보딩 (registry 기반 동적 가이드)
├── browser.py # Dev.to 브라우저 자동화
│
│ ── EventBus 플러그인 (독립 모듈) ──
│
├── safety.py # Rate limiting + 콘텐츠 검증 (strike.before 차단)
├── memory.py # SQLite 영속 저장소 (활동 이력 + 중복 방지 + campaign_id/utm_url)
├── tracker.py # 답글 추적 (댓글 남긴 게시글의 새 답글 감지)
├── conversion.py # UTM 삽입 + 전환 기록 (strike.before/after)
├── scheduler.py # 예약 발행 (시간 도래 → 자동 실행)
│
│ ── 독립 모듈 ──
│
├── campaign.py # Campaign CRUD + KPI 리포트
├── asset.py # Asset library (에셋 저장/검색/재사용)
├── message.py # MessageFrame 관리 (ICP별 메시지 프레임)
├── measure.py # Attribution + A/B 실험 + weekly report
├── strategy.py # 주간 플랜 자동 생성 + 저위험 자동 스케줄
├── monitor.py # 모니터링 데이터 집계 (SQLite → JSON)
├── dashboard.py # 대시보드 웹서버 (aiohttp, 별도 프로세스)
├── static/index.html # 대시보드 프론트엔드 (단일 HTML, vanilla JS)
├── persona.py # 플랫폼별 페르소나 관리 (파일 I/O만)
├── llm.py # 내장 LLM 댓글 생성 (anthropic SDK)
├── autonomous.py # 자율 루프 엔진 (pipeline + events + scheduler 통합)
├── daemon.py # gwanjong-daemon CLI 진입점
run.py # 직접 실행 진입점
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 277 lines · 3,295 tokens per session scan A 483457c87a89
gwanjong-mcp CLAUDE.md is an instructions file published in the GitHub repository SonAIengine/gwanjong-mcp (0 stars, last pushed 5mo ago), licensed MIT. It adds 3,295 tokens to every session, about $0.0165 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
forem CLAUDE.md
Claude Code instructions for forem/forem, a project described as: For empowering community 🌱.
rustchain-bounties claims.instructions.md
Instructions for Scottcjn/rustchain-bounties: Claim files are public audit records for bounty submissions and code reviews.
valuein CLAUDE.md
Claude Code instructions for valuein/valuein, covering claude.md, what this repo is (and isn't), sibling repos — where the actual code lives, commands and with a paid/sp500 token.
g7 AGENTS.md
AGENTS.md instructions for gnuboard/g7, covering 그누보드7 development guide, 빠른 참조 - 상세 가이드 문서, 백엔드 backend/ (34개), 프론트엔드 frontend/ (50개) and 확장 시스템 extension/ (30개).
korean-firefighter-law-mcp CLAUDE.md
Claude Code instructions for ssd7830-cmyk/korean-firefighter-law-mcp, covering korean-firefighter-law-mcp, a. 답변 규칙 — 기억이 아니라 이 저장소의 코드가 근거다, b. mcp 도구 사용 and c. 작업 규칙.
lupa-mcp CLAUDE.md
Claude Code instructions for paeyoungpark-web/lupa-mcp, covering claude.md — lupa-mcp 개발 노트, 무엇인가, 설계 규칙 (어기지 말 것), 알아둘 사실 and 상태.