Claude configuration

Subsystems

Folder Contents Pages
hooks/ 119 registered hook commands over 70 scripts, plus other hook modules 99
skills/ 82 SKILL.md skills in 12 sources 82
commands/ 14 slash commands 14
tools/ Python CLI utilities and the session-management package 22

Hook lifecycle

Claude Code hook lifecycle by event and matcher

Hook types in a $HOME-rooted session are counted on the hooks index.

Where skills load from

Which sessions load which skill sources

Notes

This section documents the Claude Code configuration on this host from the live files: every hook command, every skill and every slash command has its own page, written from the registrations and the scripts themselves and refreshed by scripts/gen-claude-pages.py in this repository. The generator reads ~/.claude/settings.json and settings.local.json, the installed plugins' hooks.json (orchestrata 1.5.0, tmux-status 1.2.0, turn-navigator 1.1.3, codetools 0.1.0), the worktree's repo-scoped settings, every SKILL.md and the commands directory, and rewrites only the generated parts of each page: the hand-written text between the prose markers is kept byte for byte.

How the pieces fit. Hooks are the mechanism: a script runs at an event and can inject context, run a side effect, ask, or deny. Skills are instruction packs the model loads when their description matches. Commands are instruction files the user invokes by name. The rulings (RUL-014: a ruling needs a mechanism) are what most hooks enforce, so the page for a guard links the ruling it implements. The policy side lives in workflow, gates and briefs, the project in Orchestrata Ultimate and the tooling in codetools. The session tools under tools (the session CLI and the file tools) are what the session hooks and commands call.

Reading order for a newcomer: the hook lifecycle to see what fires when; the shell-guards skill to learn what to do when a call is refused; then the skills and commands indexes for what is available and what is dead on this host.

Earlier documentation

Documented from the orchestrata-ultimate-architecture repository: .claude/skills/, .claude/hooks/, and claude/ (tools + Python hooks). Every page carries the full source, a per-function reference (what each function does and its parameters), and a zoomable diagram — click any diagram to open the full-size SVG in a new tab.

Subsystems

Folder Contents Pages
skills/ Agent-invocable capability packs (SKILL.md) 4
tools/ Python CLI utilities and the session-management package 22
hooks/ Event-driven guards and injectors (shell + Python) 48

How the pieces fit

Diagram claude-index-1

Notable findings from documentation

Documented verbatim, not fixed (source defects flagged on the relevant pages):