dotclaude

Adds a lean output style, six role agents, and a status line command. Adds hooks that ask before work on a repo with an AI policy, check edits and `git` calls, bound the context of subagents, and check Markdown line breaks and drift.

10 skills · 6 agents

Measures the start of a Claude Code session by source, and the usage of each run: model calls, context that each call reads again, and skills that a period did not use.

2 skills

Drives a browser with `agent-browser`, or with CloakBrowser on sites with bot detection. Answers each `WebSearch` call with a search in the headless browser. With the `cloakbrowser` backend, `WebSearch` needs the CloakBrowser binary.

1 skill

Gives Claude the code graph of a project from the CodeGraph CLI `codegraph`. Syncs the index at session start, after an edit, and after a `git` command that changes the tree. Adds one line about the graph commands at session start.

hooks

Gets current library and API docs with the Context7 CLI `ctx7`. Works with no key.

1 skill

Tools for plugin authors. Runs Claude Code in a sandbox, captures and sizes the requests, writes the JSON Schema of `settings.json`, and compares the eval scores of two plugin folders.

1 skill

Loads `AGENTS.md` files where Claude Code loads `CLAUDE.md`: the user folder, each ancestor of the working directory, nested folders on each read or edit, and the folders of `permissions.additionalDirectories`. Skips a file that a `CLAUDE` file already loads.

Shows one thing that you should know before Claude changes something, in a band above the prompt that does not stop the change. Opt-in. Makes at most one model call per turn: a Haiku call, or a fork of the conversation when the Haiku call gets HTTP 429.