[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-1208":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":10,"rankLanguage":10,"license":22,"archived":23,"fork":23,"defaultBranch":24,"hasWiki":25,"hasPages":25,"topics":26,"createdAt":10,"pushedAt":10,"updatedAt":47,"readmeContent":48,"aiSummary":49,"trendingCount":16,"starSnapshotCount":16,"syncStatus":50,"lastSyncTime":51,"discoverSource":52},1208,"ctx","stevesolun\u002Fctx","stevesolun","Skill, agent, MCP, and harness recommendations for Claude Code\u002Fcustom LLMs: 102,928-node LLM-wiki graph, 91,464 skills, 467 agents, 10,790 MCPs, 207 harnesses, and capped execution recommendations.","https:\u002F\u002Fstevesolun.github.io\u002Fctx\u002F",null,"Python",483,58,4,1,0,33,110,177,99,5.31,"MIT License",false,"main",true,[27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46],"agents","ai-agents","anthropic","automation","claude","claude-code","context-management","developer-tools","harness","knowledge-graph","llm","llm-wiki","mcp","micro-skills","obsidian","real-time","recommendation-engine","skill-management","skills","wiki","2026-06-12 02:00:24","# ctx — Skill, Agent, MCP & Harness Recommendations\n\n[![License: MIT](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-MIT-yellow.svg)](LICENSE)\n[![Python 3.11+](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPython-3.11+-green.svg)](https:\u002F\u002Fpython.org)\n[![PyPI](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fclaude-ctx.svg)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fclaude-ctx\u002F)\n[![Tests](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FTests-3645_collected-brightgreen.svg)](#)\n[![Graph](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FGraph-102%2C696_nodes_\u002F_2.9M_edges-red.svg)](graph\u002F)\n[![Docs](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdocs-MkDocs_Material-blue.svg)](https:\u002F\u002Fstevesolun.github.io\u002Fctx\u002F)\n\nctx watches what you are building, walks a **102,696-node** graph, and\nrecommends a small, top-scored bundle of skills, agents, and MCP servers for\nthe current task. If you use your own local\u002FAPI model instead of Claude Code,\nctx has a separate harness catalog flow: tell it the model and goal, review the\nrecommended harness, then install with dry-run\u002Fupdate\u002Funinstall controls.\n\nCurrent shipped snapshot:\n\n- **91,432 skills**: 1,969 curated\u002Fimported skills plus **89,463 body-backed Skills.sh skills**.\n- **464 agents**, **10,787 MCP servers**, and **13 cataloged harnesses**.\n- **2.9M graph edges** across semantic similarity, tags, slug tokens, source overlap, direct links, quality, usage, type affinity, and graph structure.\n- **89,463 hydrated `SKILL.md` bodies** in the shipped LLM-wiki; long entries are converted through the micro-skill gate instead of loading raw long prompts.\n- Entity updates for skills, agents, MCPs, and harnesses print benefits\u002Frisks and skip replacement unless you explicitly approve the update.\n\n## Why it exists\n\n- **Discovery** — with 91K+ skill nodes, 460+ agents, 10K+ MCP servers, and 13 cataloged harnesses, you can't possibly know which exist or which apply to your current work.\n- **Context budget** — loading everything wastes tokens and degrades quality. You need the right 10–15 per session.\n- **Skill rot** — skills you installed months ago and never used are cluttering context. Stale ones should be flagged automatically.\n\n## Install\n\n```bash\npip install claude-ctx\nctx-init                    # terminal wizard: hooks, graph, model, harness goal\nctx-init --wizard           # force the same wizard from scripts\u002Ftests\nctx-init --model-mode skip  # non-interactive setup for automation\nctx-init --model-mode custom --model openai\u002Fgpt-5.5 --goal \"build a CAD agent\"\n```\n\nOptional extras: `pip install \"claude-ctx[embeddings]\"` for the semantic backend, `pip install \"claude-ctx[harness]\"` for local\u002FAPI model harness runs, `pip install \"claude-ctx[dev]\"` for the test toolchain.\n\n### Pre-built knowledge graph (optional)\n\nA pre-built knowledge graph of 102,696 nodes and 2.9M edges ships as a tarball. The same tarball includes `external-catalogs\u002Fskills-sh\u002Fcatalog.json`, 89,463 body-backed Skills.sh skill pages under `entities\u002Fskills\u002Fskills-sh-*.md`, 89,463 hydrated installable Skills.sh `SKILL.md` files under `converted\u002Fskills-sh-*\u002F`, and 13 cataloged harness pages under `entities\u002Fharnesses\u002F`. Extract to get a ready-to-use `~\u002F.claude\u002Fskill-wiki\u002F`:\n\n```bash\n# after `git clone` — or download graph\u002Fwiki-graph.tar.gz from the GitHub release\nmkdir -p ~\u002F.claude\u002Fskill-wiki\ntar xzf graph\u002Fwiki-graph.tar.gz -C ~\u002F.claude\u002Fskill-wiki\u002F\n```\n\n> **Windows \u002F Git-Bash \u002F MSYS:** pass `--force-local` so `tar` doesn't read the `c:` in the path as a remote host: `tar --force-local xzf graph\u002Fwiki-graph.tar.gz -C ~\u002F.claude\u002Fskill-wiki\u002F`. Linux\u002FmacOS users can ignore.\n\n## Use\n\nAfter install, the `ctx` hooks integrate automatically with Claude Code's `PostToolUse` + `Stop` events. Typical flow:\n\n```bash\nctx-scan-repo --repo .     # scan current repo and stack signals\nctx-scan-repo --repo . --recommend  # include skill\u002Fagent\u002FMCP recommendations\nctx-agent-add --agent-path .\u002Fcode-reviewer.md --name code-reviewer\nctx-harness-add --repo https:\u002F\u002Fgithub.com\u002Fearthtojake\u002Ftext-to-cad --tag cad\nctx-harness-install text-to-cad --dry-run   # inspect before cloning\u002Frunning anything\nctx-harness-install text-to-cad --update --dry-run\nctx-harness-install text-to-cad --uninstall --dry-run\nctx-skill-quality list     # four-signal quality score for every skill\nctx-skill-quality explain python-patterns   # drill into a single skill\nctx-skill-health dashboard # structural health + drift detection\nctx-toolbox run --event pre-commit          # run a council on the current diff\nctx-monitor serve          # local dashboard: http:\u002F\u002F127.0.0.1:8765\u002F\n```\n\nThe **`ctx-monitor`** dashboard shows currently loaded skills, agents, MCP servers, and installed harness records. It provides load\u002Funload buttons where ctx owns the live action, a cytoscape graph view (`\u002Fgraph?slug=…`), the LLM-wiki entity browser (`\u002Fwiki\u002F\u003Cslug>`), a filterable skills grid, a session timeline, an audit log viewer, and a live SSE event stream. Installed harness records appear in `\u002Floaded`; cataloged harnesses appear in `\u002Fwiki` and `\u002Fgraph`. Harness install\u002Fupdate\u002Funinstall actions stay in `ctx-harness-install`.\n\nWhen `ctx-skill-add`, `ctx-agent-add`, `ctx-mcp-add`, or `ctx-harness-add`\nfinds an existing entity, ctx prints a benefits\u002Frisks update review and skips\nreplacement by default. Re-run with `--update-existing` to apply the catalog or\nlocal asset update after review.\n\nStep-by-step entity onboarding:\n**\u003Chttps:\u002F\u002Fstevesolun.github.io\u002Fctx\u002Fentity-onboarding\u002F>**\n\nFull docs, architecture, and every module: **\u003Chttps:\u002F\u002Fstevesolun.github.io\u002Fctx\u002F>**\n\n## License\n\nMIT — see [LICENSE](LICENSE).\n","ctx 是一个为 Claude Code 或自定义 LLMs 提供技能、代理、MCP 和执行框架推荐的工具，内置了一个包含 102,696 节点和 2.9 百万条边的知识图谱。它通过分析用户当前任务，并在庞大的图谱中找到最相关的技能组合、代理及 MCP 服务器进行推荐，帮助开发者高效管理上下文预算，避免因加载过多信息而浪费资源或降低性能。此外，ctx 还能自动识别并标记不再使用的过时技能，确保系统始终保持最佳状态。适用于需要自动化处理复杂 AI 任务或者希望优化其 LLM 应用程序性能的开发人员。",2,"2026-06-11 02:42:19","CREATED_QUERY"]