freemcplab / Directory / Playground / Everything Claude Code

● Community affaan-m ⚡ Instant

Everything Claude Code

Name: Everything Claude Code (Claude Skill)
Author: affaan-m

by affaan-m · affaan-m/everything-claude-code

A complete agent-harness performance kit — opinionated skills, instincts, memory, and security patterns to make any coding agent reliable.

everything-claude-code is a curated bundle that codifies what the author considers production-ready agent practice: research-first development, persistent memory patterns, security guardrails, and 'instincts' (heuristics agents should default to). Works with Claude Code, Codex, OpenCode, Cursor — anywhere skills load. Treat it as a paved-road config you can fork and trim.

▶ Watch demo 📦 Install 💡 Use Cases

Why use it

Key features

Research-first development discipline (prove understanding before writing)
Persistent memory + handoff patterns
Security defaults (no shell unless asked, no creds in chat)
Multi-harness compatible (Claude Code, Codex, OpenCode, Cursor)
Active maintenance, large community

Live Demo

What it looks like in practice

ready

Install

Author repository · Claude Skills documentation

Use Cases

Real-world ways to use Everything Claude Code

Bootstrap a new project with strong defaults for any coding agent

👤 Tech leads setting team standards ⏱ ~30 min intermediate

When to use: You're starting a project and you want every agent (yours and your team's) to behave the same way.

Flow

Install

Clone everything-claude-code. List the skill modules and their purposes.✓ Copied

→ Module inventory with descriptions
Trim

Disable modules I don't need (e.g., crypto). Keep research-first, memory, security.✓ Copied

→ Pruned skill set
Validate

Run a sample prompt that should trigger research-first behavior. Confirm it does.✓ Copied

→ Agent gathers context before writing

Outcome: Every agent on the project starts with the same opinionated defaults.

Pitfalls

Skill conflicts with team's existing CLAUDE.md — Order matters — project CLAUDE.md takes precedence; reconcile explicitly

Combine with: filesystem

Add security and memory patterns to an existing setup

👤 Devs who've been improvising agent config ⏱ ~45 min intermediate

When to use: Your agent has been useful but inconsistent — sometimes shell-happy, sometimes forgetful.

Flow

Audit

Compare my current CLAUDE.md vs everything-claude-code's recommended skeleton. Highlight gaps.✓ Copied

→ Gap list with rationale
Adopt

Pull in the security and memory modules only. Diff what would change.✓ Copied

→ Diff to review

Outcome: Same agent, more reliable behavior.

Pitfalls

Adopting the whole kit changes too much at once — Per-module adoption; verify each before next

Combinations

Pair with other MCPs for X10 leverage

everything-claude-code-skill + git-mcp-idosal

Research-first + always-fresh repo source

Research-first via everything-claude-code; pull library source via GitMCP.✓ Copied

everything-claude-code-skill + claude-best-practice-skill

Layer two opinionated skill kits with conflict resolution

Adopt shanraisshan/claude-code-best-practice on top; resolve conflicts in CLAUDE.md.✓ Copied

Tools

What this MCP exposes

Tool	Inputs	When to call	Cost
research_first	task description	Default — before changes	tokens for the research step
memory_handoff	session summary	End of session	tokens
security_lint	diff	Before commit	tokens

Cost & Limits

What this costs to run

API quota: N/A
Tokens per call: Higher per-task (research adds context) but fewer total iterations
Monetary: Free
Tip: Adopt incrementally — measure before/after token spend per task type

Security

Permissions, secrets, blast radius

Minimum scopes: Read access to project workspace

Credential storage: Per-host (the skill itself stores no creds)

Data egress: None (skill is local config)

Big bundles can pull behavior surprises — review modules you enable

Troubleshooting

Common errors and fixes

Conflicting CLAUDE.md guidance

Resolve precedence — project > user > skill

Token spend went up

Disable modules you don't need; research-first is heaviest

Alternatives