Project guidance lives in AGENTS.md. Read it first. This file holds only what is specific to Claude Code, so the two never drift apart — duplicating commands or architecture across both is how they start disagreeing. - Plan before multi-file edits; build one small change at a time. - Use the exact verification commands in agent_docs/testing.md — don't invent package-manager scripts. - Delegate focused research, review, or debugging to subagents. - Never auto-approve MCP, shell/write/network, production, billing, or destructive tools.…
A standalone agent skill for reviewing and improving interface polish, published for installation via npx skills add jakubkrehel/make-interfaces-feel-better. It is documentation-only; there is no build, lint, or test tooling. The skill lives in skills/make-interfaces-feel-better/:
This file provides context for Claude Code when working with this codebase. claudecode.nvim - A Neovim plugin that implements the same WebSocket-based MCP protocol as Anthropic's official IDE extensions. Built with pure Lua and zero dependencies. Best Practice: Always use mise run all at the end of editing sessions for complete validation. The dev toolchain is provisioned by mise (see mise.toml), which replaced the former Nix flake devShell. The fixtures/ directory contains test Neovim configurations for verifying plugin integrations: Available…
Follow .github/workflows/tests.yml for the canonical sequence. Tests take ~8 minutes. Always save output to a file and grep from there: Before committing, run dotnet format ./src/ -v:diag and fix style issues. Semantic commit messages: label(scope): description Labels: fix, feat, chore, docs, test, devops Never add Co-Authored-By agents in commit message. Never add "Generated with" in commit message. Never add test plan to PR description. Keep PR description short — a few bullet points at most. Branch naming for issue fixes:…
Quality over speed. Small steps compound into big progress. When choosing the next task, prioritize in this order: Fail fast on risky work. Save easy wins for later. If yes,…
For agents working in this repo, commerce-builder plugin users included. The public reference for commerce agents on Claude: a shopping agent and a merchant agent on three paths each, four vertical examples, and a Claude Code plugin. requirements.txt installs the seven packages and their pinned dependencies (requirements-dev.txt adds pytest and ruff); scripts/install.sh runs it. No real company, brand, product, or person appears: the only company is ACME and its lines; every brand, prompt, schema, and figure is invented here. Two…
This repository ships a four-layer marketing operating system: 112 discipline skills, 8 protocol skills, and 8 command entrypoints. This file is a compact navigation and safety map, not a second handbook. The authoritative topology is references/system-catalog.json; its generated human view is docs/system-architecture.md. Do not preload every linked document. Start with this map, select the smallest route, then load the chosen Skill and only its explicit runtime dependencies. Controllers keep schemas, hashes, permissions, and validation outside model-visible prose. Canonical order: Narrative…
Canonical agent instructions for /Users/kaitran/CloudPersonal/ccs/cli. AGENTS.md must stay a symlink to this file. CCS is a TypeScript/Bun CLI and dashboard for managing Claude Code, Codex, Factory Droid, CLIProxy, and compatible provider profiles. - Default branch is dev. Feature/fix branches start from dev; production hotfixes start from main only when explicitly needed. - Never touch the user's real ~/.ccs/ or ~/.claude/ in tests. Use getCcsDir() from src/utils/config-manager.ts; it respects CCSHOME. - Do not commit directly to dev or main. - Do…
Always read and follow AGENT.md — unconditionally, in full. It holds the general, durable insights about this product (architecture, warehouse backends, cross-cutting design principles). Treat them as binding context for any change here. Keep general product insights in AGENT.md (not here); this file is only for Claude-specific notes.
Layout: pnpm + turbo monorepo — core/ (gateway, agent, runtime, identity, meetings), clients/terminal (Next.js 15, port 3000, npm run dev), docs/docs (Mintlify site — the published law), calm/ (FINOS CALM model). Licensing is FINOS-gated: new deps must be Category A (MIT/BSD/Apache); weak-copyleft needs an entry in license-exceptions.json (ADR-0004) — never add GPL/AGPL.
"Claude Code from Source" — a technical book analyzing the architecture of Anthropic's Claude Code CLI, reverse-engineered from source maps leaked via npm. Written in the style of an O'Reilly…
CompozyOS is a Go daemon managing ACP agents over JSON-RPC/stdio, SQLite state, HTTP/SSE, UDS, and a web UI. Capabilities must be extensible and manageable by agents through structured CLI/HTTP/UDS surfaces. Real users run every release. Preserve the three regimes from SD-013 and L-040: Read the relevant subtree instructions before working there: Load a skill when the user names it or its procedure resolves a task-specific need; start with the owning skill.
Applies to internal/ and cmd/compozy. Root CLAUDE.md owns compatibility, delivery, and test placement. Use eng-code-guidelines for project Go conventions and eng-test-conventions when editing Go tests; load other references for the particular problem.
React SPA using Vite, TanStack Router/Query, Tailwind, shadcn/ui, XState Store, and Zod. Root CLAUDE.md owns compatibility and delivery; this file adds web conventions. Use app-renderer-systems for a new or restructured domain system; eng-data-boundaries for changed authorization, pagination, cache, or stream contracts; xstate-store for store lifecycle changes; react for React-specific patterns. Load other library references only for the API or pattern in question. Use eng-design for design-system/redesign work, eng-ui-screenshot for named-reference comparison, and storybook-stories for stories. Run from the repository root;…
This document outlines best practices for working with Claude Code to ensure efficient and successful software development tasks. For complex or multi-step tasks, Claude Code will use: * TodoWrite: To create a structured task list, breaking down the work into manageable steps. This provides clarity on the plan and allows for tracking progress. * TodoRead: To review the current list of tasks and their status, ensuring alignment and that all objectives are being addressed. Understanding file content is crucial before…
Use this agent when: - User asks to find specific emails or types of emails - Complex multi-criteria email searches are needed - Need to discover emails through progressive refinement…
A file Claude Code reads at the start of every session. It holds the commands, conventions and warnings the agent needs for this project.
Where does it go?
At the repository root. Claude Code also reads CLAUDE.md files in subdirectories when it works there.
What should it contain?
Build and test commands, the project's layout, conventions that aren't obvious from the code, and mistakes to avoid. Short files tend to work better than long ones.
CLAUDE.md or AGENTS.md?
Claude Code reads CLAUDE.md; most other agents read AGENTS.md. Many projects keep one and point the other at it.
KhazPCLAUDE.md
