Why Everything AI Coding?
AI coding agents are improving fast, but the ecosystem around them is still fragmented. Finding a reliable MCP server, reusable skill, practical rule set, or prompt collection usually means searching across multiple repositories and formats.
Everything AI Coding is a curated knowledge base that continuously collects, deduplicates, enriches, and scores resources from 9+ upstream sources. Every entry includes quality signals — LLM-scored coding relevance, documentation quality, specificity, plus health metrics like freshness and community popularity — so you can evaluate before you install. Browse right here on GitHub, explore interactively with the web catalog, or search and install with one command using the Coding Hub tool.
📚 Knowledge Base
🔌 MCP Servers — 5117 entries
Model Context Protocol servers that connect AI agents to external tools, databases, and services.
Browse all MCP servers → · Browse interactively →
🎯 Skills — 24886 entries
Reusable agent capabilities and workflows for AI coding assistants.
Browse all skills → · Browse interactively →
📋 Rules — 245 entries
Coding conventions and AI behavior guidelines for consistent development.
Browse all rules → · Browse interactively →
💡 Prompts — 876 entries
Developer-focused prompt templates for common coding tasks.
Browse all prompts → · Browse interactively →
🧩 Plugins — 3374 entries
Bundled marketplace plugins (skills + commands + agents + MCP servers).
Primarily for Claude Code; opencode partially compatible (npm); cursor / windsurf / costrict have no equivalent mechanism.
Browse all plugins → · Browse interactively →
Private-deployment mirror: the plugin catalog is the upstream data source for costrict-plugin-marketplace, which bundles every verified plugin as bare git repos for customers running
cscin air-gapped networks. See that project's README for the bundle format and the clientimport.shflow.
Data Sources & Quality
Everything AI Coding aggregates data from multiple upstream sources, then enriches, scores, and republishes the cleaned catalog.
Quality scoring
Every entry is scored on a 0–100 composite scale: final_score = LLM × 0.85 + health × 0.15.
LLM dimensions (1–5, up to 6 per type): coding relevance, doc completeness, description accuracy, writing quality, specificity, install clarity (MCP & Skills only)
Health signals: freshness (🟢 Active / 🟡 Stale / 🔴 Abandoned), popularity (GitHub stars), source trust (upstream reputation)
Decisions: accept (≥ 65) · review (50–64) · reject (< 50)
Each sub-directory README shows the Top 100 entries ranked by this composite score.
Pipeline
- Sync —
scripts/sync_*.pypull from upstream sources weekly - Merge —
scripts/merge_index.pydeduplicates across sources, merges metadata - Evaluate — Single LLM call: 6-dimension scoring + enrichment (tags, summary, tech_stack) + health signals
- Download —
scripts/download_catalog.pyfetches per-entry files intocatalog-download/and reconcilescatalog/index.jsonwith what actually landed (docs) - Bundle —
scripts/build_catalog_bundle.pypacksindex.json+catalog-download/intodist/catalog-bundle.tar.gzfor downstream consumers (docs) - Publish — GitHub Actions refresh the catalog, generate README tables, and update the web catalog
everything-ai-coding/
├── install.sh / install.ps1 # One-command installers
├── catalog/ # Generated catalog data
│ ├── index.json # Full merged index (4000+ entries)
│ ├── search-index.json # Lightweight search index
│ ├── mcp/ # MCP servers — index + README
│ ├── skills/ # Skills — index + README
│ ├── rules/ # Rules — index + README
│ └── prompts/ # Prompts — index + README
├── platforms/ # Per-platform skill + command definitions
├── scripts/ # Sync, merge, scoring, and generation scripts
└── .github/workflows/ # CI automation
🛠 Coding Hub — Search & Install
Everything AI Coding also ships as a CLI tool that lets you search, browse, and install resources directly from your AI coding agent.
Quick Start
Install for your platform with one command:
macOS / Linux
# Costrict CLI (run from your project root)
curl -fsSL https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/install.sh | bash -s -- --platform costrict
# VSCode Costrict extension
curl -fsSL https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/install.sh | bash -s -- --platform vscode-costrict
# Claude Code
curl -fsSL https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/install.sh | bash -s -- --platform claude-code
# Opencode (run from your project root)
curl -fsSL https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/install.sh | bash -s -- --platform opencode
Windows (PowerShell)
# Costrict CLI
irm https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/install.ps1 | iex
# Specify the platform manually if auto-detection fails
& ([scriptblock]::Create((irm https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/install.ps1))) -Platform costrict
After installation, try a search command:
# Claude Code
/eac:search typescript
# Opencode / Costrict CLI / VSCode Costrict (Roo Code)
/eac-search typescript
You can also hand the installation off to another agent with this prompt:
You are an installation assistant. Open the following URL, read the "For Agents" section,
and follow it exactly to install Everything AI Coding for the platform you are currently running on.
Do not clone the repository. Only read this raw file:
https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/README.md
After installation, report which platform path was used and whether verification succeeded.
Platforms
Everything AI Coding supports four AI coding platforms. The resource catalog is shared, while command names and installation paths vary slightly by platform.
- Costrict CLI installs command files per project, so run the installer in each repository root.
- VSCode Costrict / Roo Code installs command files globally under
~/.roo/commands/. - Claude Code can load the full skill and subcommands from the global skills directory.
- Opencode installs commands per project under
.opencode/command/.
For Agents
If you are an AI agent installing Everything AI Coding automatically, follow this section exactly.
Step 1: Run the installer
The install scripts auto-detect the host platform when possible.
macOS / Linux
curl -fsSL https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/install.sh | bash
Windows (PowerShell)
irm https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/install.ps1 | iex
Auto-detection uses process-level environment variables that each platform injects at startup:
If auto-detection fails, specify the platform manually.
macOS / Linux
curl -fsSL https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/install.sh | bash -s -- --platform <platform>
Windows (PowerShell)
& ([scriptblock]::Create((irm https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/install.ps1))) -Platform <platform>
Examples:
# Claude Code (macOS/Linux)
curl -fsSL https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/install.sh | bash -s -- --platform claude-code
# Opencode (run from the project root)
curl -fsSL https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/install.sh | bash -s -- --platform opencode
# Claude Code (Windows)
& ([scriptblock]::Create((irm https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/install.ps1))) -Platform claude-code
Run the platform-specific commands below if you need to download the files directly.
Costrict CLI (global skill + project commands)
mkdir -p ~/.costrict/skills/eac
curl -fsSL "https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/platforms/costrict/skills/eac/SKILL.md" -o ~/.costrict/skills/eac/SKILL.md
mkdir -p .costrict/eac/commands
for cmd in search browse recommend install uninstall update; do
curl -fsSL "https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/platforms/costrict/commands/eac/eac-${cmd}.md" -o .costrict/eac/commands/eac-${cmd}.md
done
VSCode Costrict extension / Roo Code (global skill + global commands)
mkdir -p ~/.costrict/skills/eac
curl -fsSL "https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/platforms/vscode-costrict/skills/eac/SKILL.md" -o ~/.costrict/skills/eac/SKILL.md
mkdir -p ~/.roo/commands
for cmd in search browse recommend install uninstall update; do
curl -fsSL "https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/platforms/vscode-costrict/commands/eac/eac-${cmd}.md" -o ~/.roo/commands/eac-${cmd}.md
done
Claude Code (global skill + global commands)
mkdir -p ~/.claude/skills/eac
curl -fsSL "https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/platforms/claude-code/skills/eac/SKILL.md" -o ~/.claude/skills/eac/SKILL.md
for cmd in search browse recommend install uninstall update; do
curl -fsSL "https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/platforms/claude-code/commands/eac/${cmd}.md" -o ~/.claude/skills/eac/${cmd}.md
done
Opencode (global skill + project commands)
mkdir -p ~/.opencode/skills/eac
curl -fsSL "https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/platforms/opencode/skills/eac/SKILL.md" -o ~/.opencode/skills/eac/SKILL.md
mkdir -p .opencode/command
for cmd in search browse recommend install uninstall update; do
curl -fsSL "https://raw.githubusercontent.com/zgsm-ai/everything-ai-coding/main/platforms/opencode/command/eac-${cmd}.md" -o .opencode/command/eac-${cmd}.md
done
Step 2: Verify the installation
Run a search command to confirm that the commands are available:
# Costrict CLI / Opencode
/eac-search typescript
# VSCode Costrict extension
Send this in the chat: "Search typescript with eac"
# Claude Code
/eac:search typescript
Expected result: a Markdown table of matching resources. If the command is missing, the installation did not land in the correct platform path.
Available commands
Data endpoints
Search, browse, and recommend operate on a lightweight search index. Installation fetches a per-entry API document.
Why Costrict?
If you like the Everything AI Coding workflow, try Costrict — our broader AI coding agent platform built for stronger automation and team workflows.
Contributing
Contributions are welcome. Open a PR against the appropriate directory under catalog/ and make sure the resource:
- is relevant to coding or AI-assisted development,
- has an accurate
source_url,description, and tags, - follows
catalog/schema.json.
If you need maintainers' context instead of quick-start instructions, see the repository wiki in docs/wiki/.
Acknowledgments
The /eac:evo command's quality rubric is adapted from darwin-skill by 花叔 (MIT License) — an autonomous skill optimization system inspired by Karpathy's autoresearch. darwin-skill first systematized a "structure + effectiveness" dual-evaluation framework for SKILL.md files with a ratchet mechanism for keeping only measurable improvements. Everything AI Coding simplifies it for on-demand client-side use: the dynamic live-testing dimension and the ratchet are deferred; the core dimensions (workflow clarity, instruction specificity, edge-case coverage, checkpoint design, overall architecture, etc.) are preserved and rewritten for our on-demand evo context. See docs/wiki/evo-rubric.md for the full rubric specification.
Thanks to 花叔 for making the ideas and the original skill publicly available.
Disclaimer
Everything AI Coding is an index and installation helper for third-party open-source resources. Every MCP server, skill, rule, and prompt listed in the catalog remains the property of its original author.
This repository does not guarantee the safety, availability, accuracy, or compliance of third-party resources. Review source code and licenses before use, and open an issue if you find security or copyright problems.
Everything AI Coding is released under the MIT License.
