Claude Code Artifacts

A public library of reusable Claude Code artifacts — skills, hooks, rules, environment configurations, agents, and slash commands — ready to drop into any Claude Code setup.

Each artifact is a self-contained building block for Claude Code. You can consume it two ways: reference it from a single environment YAML file and let the Claude Code Toolbox install it for you (by raw URL or a shared base-url), or copy a rule, a hook, or a skill straight into your ~/.claude/ directory. Ready-made environment configurations take the first way further: one YAML file wires several artifacts into a complete setup the toolbox installs in one command.

The sections below are a catalog of what is available today, grouped by artifact type. Each entry links to the artifact itself and to that type's README for conventions and installation details.

Environments

Environment configurations are single YAML files that declare a complete Claude Code setup — artifacts, MCP servers, settings, dependencies, and more — installed in one command by the Claude Code Toolbox. They live under environments/library/; see environments/README.md for conventions, validation, and how to consume one.

serena.yaml

Gives any Claude Code setup IDE-grade code intelligence in one install: registers the Serena MCP server in a curated LSP-only mode — semantic symbol navigation (find_symbol, find_referencing_symbols, get_symbols_overview), LSP diagnostics, and symbol-aware editing (rename_symbol, replace_symbol_body, safe_delete_symbol) — while Serena's file I/O, shell, memory, onboarding, and dashboard tools stay disabled so they never compete with Claude Code's built-ins. The configuration is composed of supporting files from this repository that work as parts of this setup rather than on their own: the serena-tool-selection skill, which routes code navigation to the semantic tools and documents their known limitations with workarounds; two Search/Grep steering tiers — the advisory serena_tool_enforcement.py command hook with its companion configuration, plus an enforcing LLM prompt gate defined inline in the YAML; and the Serena runtime files lsp-only.yml and system_prompt.yml deployed to ~/.serena/. See environments/library/serena.yaml.

Skills

Agent Skills are multi-file packages that extend Claude Code with specialized, progressively disclosed capabilities. They live under skills/library/; see skills/README.md for the layout, conventions, and the two ways to install one.

dynamic-workflow-patterns

Pattern taxonomy and orchestration discipline for Claude Code dynamic workflows: which of the six workflow patterns fits a task, which agent roles to combine for each task family, which model to route to each role, and how to keep a run alive through server errors (for example HTTP 529), recover interrupted runs, and honor token budgets. Load it before authoring or running a Workflow script rather than hand-rolling one from memory. See skills/library/dynamic-workflow-patterns/SKILL.md.

Hooks

Hooks run custom logic at Claude Code lifecycle events such as PreToolUse, PostToolUse, UserPromptSubmit, and SessionStart. Scripts live under hooks/library/; see hooks/README.md for conventions and how to wire a hook to an event.

idle_notification.py

Sends a desktop notification when Claude Code goes idle — waiting for your input after roughly 60 seconds of inactivity — so you can step away and be pulled back when Claude needs you. Wires to the Notification event with the idle_prompt matcher and falls back across notification backends per platform (desktop notifier, then a platform command-line fallback). See hooks/library/idle_notification.py.

status_line.py

Renders a colored Claude Code status line from the session JSON on stdin: model name, project directory, git branch (with main/master flagged in red), session id, added and removed line counts, compact 5h/7d rate-limit usage, an update-available indicator, and an optional custom suffix. Bracketed segments appear only when enabled and present. See hooks/library/status_line.py.

Both hooks read their optional YAML configuration through the shared hooks/library/hook_config_loader.py helper, so install it alongside whichever hook you use.

Using an artifact

Every artifact is a plain file addressable by its raw URL:

https://raw.githubusercontent.com/alex-feel/claude-code-artifacts-public/main/<path>

The Claude Code Toolbox installs artifacts from a single environment YAML file. For example, wiring the idle-notification hook to the Notification event — the shared helper module travels through files-to-download, because every hooks.files entry must be referenced by a hook event:

files-to-download:
  - source: "https://raw.githubusercontent.com/alex-feel/claude-code-artifacts-public/main/hooks/library/hook_config_loader.py"
    dest: "~/.claude/hooks/hook_config_loader.py"

hooks:
  files:
    - "https://raw.githubusercontent.com/alex-feel/claude-code-artifacts-public/main/hooks/library/idle_notification.py"
  events:
    - event: "Notification"
      matcher: "idle_prompt"
      type: "command"
      command: "idle_notification.py"

Skills install through the toolbox skills key or the skills CLI; each type's README shows the exact shape. An environment configuration from environments/library/ installs directly: point the toolbox setup-environment command at its raw URL, or compose it into your own configuration through the inherit key. See the toolbox Environment Configuration Guide for the full YAML reference (skills, hooks, MCP servers, settings, inheritance, and more).

Adding a new artifact

Read CONTRIBUTING.md for the full workflow. In short: place the artifact under the matching <type>/library/ directory following that type's README, update this README's catalog (see CLAUDE.md), run the quality gate locally (uv run pre-commit run --all-files), and open a pull request. Environment configurations are additionally schema-validated in CI.

Security

Please report vulnerabilities privately through GitHub Security Advisories. See SECURITY.md. Because artifacts can include executable hooks and setup commands, always review an artifact before installing it.

License

Released under the MIT License.