Caution
50% useful tool/teaching resource for agentic coding, 50% externalised temper tantrum.
Goblin Mode is a permanently-in-flux configuration system (skills, agents, hooks) built by one developer to solve problems they actually had. Every skill traces back to a specific friction point; the hooks enforce habits that were being skipped.
Note
This setup is specific to one developer's workflow (and brain). If you're building your own, the process transfers; the specifics should be yours. Start with Building Your Own.
| Component | Count | What it does |
|---|---|---|
| Skills (command) | 29 | Slash commands you invoke (e.g. /commit-one) |
| Skills (role) | 16 | Ambient knowledge that loads automatically when relevant |
| Skills (model-invocable command) | 25 | Command skills the model can also self-invoke |
| Agents | 9 | Autonomous sub-processes for multi-step work |
| Hooks | 5 global + 2 project-level | Scripts triggered by git and session events |
These counts are written by gen-skills-index.py, not hand-maintained: the skill rows come straight from its own classification, so they cannot drift from skills/README.md the way they used to. Agents and hooks are counted the same way at generation time. See Keeping this wiki honest.
- Read CLAUDE.md, the behaviour file loaded every session: tone, code conventions, git workflow.
- Browse docs/README.md, the documentation wiki: architecture and a reference page per subsystem.
- Browse skills/README.md, the full generated index of every skill.
| Directory | What's there |
|---|---|
CLAUDE.md |
The behaviour file: technical profile, communication rules, code conventions, git workflow, security defaults. Loaded every session. |
skills/ |
Command skills (invoked with /name) and role skills (load automatically). See Skills reference. |
agents/ |
Autonomous multi-step workflows Claude delegates to. See Agents reference. |
hooks/ |
Scripts on git/session events. See Hooks reference. |
library/ |
Shared references, templates, scripts, and config examples used by skills. See Library reference. |
output-styles/ |
Persona layer; british-dev-goblin.md is set at user level in settings.json with keep-coding-instructions: true, so it sits on top of Claude Code's built-in engineering instructions rather than replacing them. Voice only: the rules live in CLAUDE.md §3. |
docs/ |
This wiki: architecture, per-subsystem reference, guides and design history. |
For the full picture (how the layers fit together, what each costs to load and the deterministic-half pattern behind several skills) see Architecture.
The full guide is at docs/guides/building-your-own.md. The short version: respect the context window, start from friction rather than speculation, encode decisions rather than preferences and let it grow one problem at a time.
| Page | Description |
|---|---|
| Documentation wiki | Full index; start here for anything beyond this README |
| Architecture | How the layers fit, context-window cost, the deterministic-half pattern |
| Skills · Agents · Hooks · Library · Configuration | Per-subsystem reference |
| Building Your Own | Transferable process for building your own setup |
| Glossary | Vocabulary specific to this config |