Everything Claude Code: 182K Stars, 232 Skills, and What It Means for AI Agent Builders
A open-source project with 182,000+ GitHub stars quietly released 232 production-grade skills and 60 specialized agents for AI coding assistants.
It’s called Everything Claude Code (ECC), built by Affaan Mohammedi. And it’s the most complete blueprint for how AI agents should actually work in production.
Here’s what’s inside — and what every AI agent builder should know.
What Is Everything Claude Code?
ECC is a skills-and-agents framework for Claude Code (adaptable to Hermes, Cursor, and other agent harnesses). It installs via:
git clone https://github.com/affaan-m/everything-claude-code
cp -r .agents/skills/* ~/.hermes/skills/The framework has three layers:
| Layer | Count | Purpose |
|---|---|---|
| Skills | 232 | Task-specific workflows with quality gates |
| Agents | 60 | Specialized subagents for specific jobs |
| Hooks | Event-based | Trigger actions on lifecycle events |
Source: https://github.com/affaan-m/everything-claude-code (182K+ stars, 232 skills, 60 agents — as of May 2026)
The 232 Skills — Organized by Function
Skills are organized into categories. The ones most relevant to AI agent builders:
Engineering (78 skills):
test-driven-development— RED-GREEN-REFACTOR with LLM-as-judge [1]spec-driven-development— Write the PRD before any code [1]incremental-implementation— Thin vertical slices, feature flags [1]source-driven-development— Every claim backed by official docs [1]context-engineering— The 5-level hierarchy for agent context [1]
Content & Business (8 skills):
article-writing— Long-form content with voice matching [2]content-engine— Platform-native content systems [2]market-research— TAM/SAM/SOM, competitive analysis [2]brand-voice— Voice profile extraction from real examples [2]
Operations (12 skills):
incident-response— Post-mortem automationdeployment-pipeline— Safe, rollback-friendly deploys
Source: https://github.com/affaan-m/ECC/tree/main/skills
The 60 Agents — Specialized Subagents
The most interesting agents:
- chief-of-staff — Communication triage, inbox zero for agents
- loop-operator — Autonomous loop execution with termination conditions
- harness-optimizer — Measures and improves agent harness quality
- code-reviewer — P0/P1/P2 severity levels, citation requirements
- spec-writer — PRD generation from rough requirements
Source: https://github.com/affaan-m/everything-claude-code/tree/main/.agents
Hooks and Contexts — The Secret Sauce
Skills aren’t static markdown files. They’re loaded dynamically through contexts — the system prompt you inject changes based on which skill is active.
Hooks fire on lifecycle events:
pre-commit → run tests
post-publish → verify URL, fire webhooks
on-error → log, notify, retryThis is the same pattern we use with event-hooks.py across our 6-blog empire. ECC formalizes it.
Source: https://github.com/affaan-m/everything-claude-code/blob/main/skills/configure-ecc/SKILL.md
What ECC Gets Right (That Most Frameworks Miss)
1. Anti-rationalization rules built in
Every skill includes a section on “incorrect thoughts agents must ignore”:
- “This is too small for a full lifecycle”
- “I’ll write the spec after the code”
- “The code is self-documenting”
Our own AGENTS.md files have similar rules. ECC bakes them into every skill.
Source: https://github.com/affaan-m/agent-skills — anti-rationalization patterns adapted by multiple frameworks
2. Source-driven development as default
The source-driven-development skill requires every framework-specific code decision to be backed by official docs — not memory. For content, this means every stat needs a primary source.
3. Context engineering hierarchy
The context-engineering skill formalizes the context loading order:
- Rules files (AGENTS.md, CLAUDE.md)
- Specs and architecture docs
- Relevant source files
- Error output from last run
- Conversation history
Source: https://github.com/affaan-m/ECC/blob/main/skills/context-engineering/SKILL.md
4. Market research with decision orientation
Research outputs must end with a recommendation, not just a summary. Every claim sourced. Contrarian evidence included.
Source: https://github.com/affaan-m/ECC/blob/main/skills/market-research/SKILL.md
How We’re Using ECC Patterns
We’ve already adapted several ECC patterns:
- SDD principles in our content-pipeline skill [3]
- Context engineering hierarchy in our cron prompts [3]
- Anti-rationalization rules in our AGENTS.md files [3]
- Market research framework for weekly niche research [3]
The result: more disciplined content crons, fewer fabricated stats, and better-structured research outputs.
Bottom Line
ECC is the most complete open-source agent framework available. 232 skills, 60 agents, production-grade patterns.
If you’re building AI agents — for content, for code, for operations — it’s worth studying. The patterns are framework-agnostic and immediately applicable.
Links:
- GitHub: https://github.com/affaan-m/everything-claude-code
- Skills directory: https://github.com/affaan-m/everything-claude-code/tree/main/.agents/skills
- Content skills (ECC fork): https://github.com/affaan-m/ECC/tree/main/skills
[1] https://github.com/affaan-m/everything-claude-code/tree/main/.agents/skills [2] https://github.com/affaan-m/ECC/tree/main/skills [3] Our implementations: content-pipeline SKILL.md, AGENTS.md, cron prompts — internal references
← Back to all posts
