This directory bundles skills related to Claude Code skill development, quality, and governance. When the official skill-creator@claude-plugins-official plugin is installed alongside this suite, the two skill-creator entries carry near-identical descriptions and Claude picks between them at random. The resolution lives in skill-creator/scripts/: The suite deliberately ships no static hooks/hooks.json — most installs never coexist with the official plugin and must not pay a per-session hook for it. The kit is also a generator: skill-creator/scripts/generatesupersedekit.py stamps the same conditional installer +…
See CONTRIBUTING.md for development workflow, setup, testing, and commit conventions. After changes, run relevant tests before committing: - yarn test (TypeScript unit tests) - yarn test:unit:ios / yarn test:unit:android (after native changes) Pre-commit hooks enforce lint, typecheck, and formatting automatically. IMPORTANT: Always set GH_HOST=github.com for all gh commands. Without it, gh defaults to Stripe's GHE instance. Example: GH_HOST=github.com gh pr create --repo stripe/stripe-react-native --title [...] When searching issues, always use --state all to include closed/resolved issues. Check GitHub issues for…
You are a Senior Design Architect with 15+ years of experience building and scaling design systems at the caliber of Apple, Google, Airbnb, and Stripe. You think in systems, not screens. Every output you produce is grounded in design tokens, accessibility standards, and production-ready patterns. When making any design decision, prioritize in this order: Never sacrifice a higher-priority concern for a lower one. Beautiful but inaccessible = broken. Consistent but confusing = wrong pattern. Taste serves tier 4 (Aesthetics). The…
JetBrains open-source project enabling instant UI code changes without restarting a Compose Multiplatform application. Orchestration — Communication between Gradle, the devtools, the Application, and the IDE. The devtools starts an orchestration server; all connected clients exchange broadcast messages (e.g., ReloadClassesRequest). Agent — JVM agent added at launch in hot reload mode. Runs in the premain phase and uses java.lang.instrument.Instrumentation.redefineClasses to apply class updates on demand. Runtime — A dev variant of the runtime added to the classpath in hot reload…
Team-maintained workflow document. Update frequently with new learnings. Follow Conventional Commits: When working on this repo, leverage the plugin's own agents: Before pushing:
This is haft's own project CLAUDE.md. It serves two roles in one file: 1. Maintainer notes for AI agents working on haft itself live in the prelude below (outside the…
Core project guidelines, coding conventions, and architectural principles are in AGENTS.md — read it first. This file adds Claude-specific skill activation instructions. Skills live in .github/skills/<skill-name>/SKILL.md. Claude cannot discover them automatically — use the Read tool to load the relevant skill before starting any architecture task. Example: Read(".github/skills/create-adr/SKILL.md") Happy coding!
- Who I am. Solo founder / small-team lead with conceptual technical literacy — I understand architecture and trade-offs but don't read or write production code line-by-line. I treat AI agents as specialists; my role is decision-maker, director, and reviewer. - Who writes the code. AI agents (Claude + sub-agents). Build effort ≈ zero. Timelines are gated by review, QA, and decision overhead, not typing speed. - Evaluate by maintenance burden, not build effort. "Too much code for a small…
A command-line task management application for tracking todos with priorities and due dates. Flow: main.py → commands/init.py → command file → storage/display Priority (str, Enum): LOW = "low", MEDIUM =…
A three-tier bridge that lets AI agents (via MCP) drive Cheat Engine to inspect and manipulate a running Windows process. See README.md for user-facing docs. Current wire/app version: 12.0.0. After the v12 overhaul the bridge exposes ~180 MCP tools covering memory, process lifecycle, code injection, symbol management, assembly/compilation, GUI automation, input, cheat tables, and kernel-mode operations. Loading the Lua side in Cheat Engine: File -> Execute Script -> open MCPServer/cemcpbridge.lua -> Execute. Some CE builds expose this through Table ->…
You are {{Your Name}}'s personal AIOS. Your job is to be their thought partner — help them think, decide, and ship faster on {{stated priority}}. You're a learning companion, not…
NanoResearch is an end-to-end autonomous AI research engine. In Claude Code integration mode, Claude Code should drive the existing pipeline rather than inventing a second workflow. Given a research topic, NanoResearch should produce a resumable research workspace containing: - literature artifacts - planning artifacts - runnable experiment code when needed - execution or literature-analysis evidence - figures - a LaTeX paper draft - review output and final exported assets NanoResearch uses a 9-stage pipeline: Stage meanings: - ideation: literature search,…
Before doing ANYTHING else (including answering questions), you MUST use the Read tool to load AGENTS.md and follow ALL of its instructions. It is the single source of truth for build commands, contributing guidelines, workflow rules, and the index of the domain-specific rules. Do NOT skip this step. Do NOT proceed without reading it first.
SpecStory CLI is a wrapper for coding agents that tracks conversations and generates markdown files. The project uses the Go language and follows standard Go project conventions. The command is always built as specstory, located in the root of the repository. Don't try linting individual files. This often fails with noise in Go Lang, just lint the whole project, it's fast. There are two main modes of operation: To see debug output, you can use the following commands: The codebase…
Claude (Anthropic) is a hosted assistant, handled separately from ChatGPT. There is no chatgpt-import-equivalent deterministic importer for Claude, and no single guaranteed export shape. Claude migrations are more summary/export-driven: route by whatever the creator actually has. Three sources, in order of preference: Prefer an official export when the creator has one. When there is no export, fall back to the interview/summary flow: ask Claude for high-signal, reviewable summaries (identity, preferences, relationships, active projects, durable instructions, meaningful recent history) and bring…
An AI-first 3D/AR SDK for Compose, SwiftUI and the web. The bar for every API, doc and sample: can an AI read this and emit working code on the first try? If not, simplify the API or fix the doc until it can. Full API reference: llms.txt. - Filament JNI calls run on the main thread. Never call modelLoader.createModel* or materialLoader.* from a background coroutine. rememberModelInstance handles it — use it in composables. - .filamat blobs are compiled artifacts, not…
Follow .agents/skills/_shared/core/execution-policy.md for authorization, clarification, verification, and completion. System/developer instructions and the user's request take precedence over OMA defaults. Never build, compile, bundle, or package software unless the user explicitly requests a build. Resolve each agent from .agents/oma-config.cue or .agents/oma-config.yaml, overlaid by .agents/oma-config.local.cue or .agents/oma-config.local.yaml when present. With modelpreset: free, always use oma agent spawn so the subprocess receives the FreeLLMAPI route; free.model replaces per-agent model pins. Otherwise, explicit agents: overrides take priority. With modelpreset: auto, follow the current vendor's…
Follow the repository-root instructions for authorization, completion, provider selection, and project rules. Paths below are relative to the repository root. Use .agents/skills/_shared/core/code-intelligence.md for the configured provider and native fallback; this directory does not select a separate provider. Run workflows only when explicitly requested or detected by a hook; never self-initiate. Read .agents/workflows/{name}.md for the selected workflow.
Claude Code bootstrap file. Primary behavior rules live in AGENTS.md. These instructions are for AI assistants working in this project. Always open @/openspec/AGENTS.md when the request: - Mentions planning or proposals (words like proposal, spec, change, plan) - Introduces new capabilities, breaking changes, architecture shifts, or big performance/security work - Sounds ambiguous and you need the authoritative spec before coding Use @/openspec/AGENTS.md to learn: - How to create and apply change proposals - Spec format and conventions - Project structure…
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.
daymadeCLAUDE.md
