How do I install the Spellbook skill?

Run npx skillmds@latest add pbakaus/spellbook in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Spellbook safe to use?

This skill has not completed SkillMD's automated safety review yet. Independent scanners report: SkillSpector: PASS, Skill Scanner: PASS. Capability flags: executes scripts. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Spellbook?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Spellbook free to use?

Yes. Installing skills from SkillMD is free. This skill is licensed under Apache-2.

Who published Spellbook?

pbakaus (@pbakaus) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Spellbook by pbakaus — Agent Skill

Name: Spellbook
Rating: 4.5 (4 reviews)
Author: pbakaus

Spellbook

Demo skill for "The Dark Arts of Skill Engineering." A tiny, readable scaffold that shows five dark arts in miniature: routing (this file), a script that talks back (scripts/context.mjs), a hook that fights back (scripts/hook.mjs), making sub-agents argue (reference/argue.md), and forcing divergence (reference/diverge.md). Use when the user types /spellbook <command>, or asks to argue, diverge, or set up the demo.

pbakaus Updated Sep 17, 2026

File contents

Spellbook (starter)

The workshop build: a pocket Impeccable you assemble as the talk climbs. It critiques and re-imagines sample/landing.html (a deliberately-median page) using five dark arts in miniature. It runs as-is; every art has a YOUR TURN marker where you extend it. Read it top to bottom, then make it yours.

Setup (always do this first)

DARK ART · the script talks back. Do not guess the project state. Run:

node scripts/context.mjs

Follow whatever it prints. The script reads the environment (git branch, whether a NOTES.md exists) and returns a fresh, in-thread directive. That directive outranks anything written statically below, because it is shaped to the exact situation. If it prints NO_NOTES, create NOTES.md first. If it prints READY, continue.

Commands (the router)

DARK ART · route like a frontier model. This file is a thin router. Do not inline the technique here; load the one expert reference the command needs, and nothing else, so context stays cheap.

Command	Load this, then follow it
`/spellbook argue [page]`	`reference/argue.md`
`/spellbook diverge [page]`	`reference/diverge.md`

[page] defaults to sample/landing.html. If no command is given, ask which one the user wants. Never run both references at once.

The hook (already working, even when this skill is not invoked)

DARK ART · hooks that fight back. scripts/hook.mjs runs after every file edit (wire it up via .claude/settings.json, see README.md). It scans the touched file for a banned pattern and injects a system-reminder. You do not invoke it; it watches. Try editing a file to include TODO: fix later and watch it push back.

---name: spellbookdescription: Demo skill for "The Dark Arts of Skill Engineering." A tiny, readable scaffold that shows five dark arts in miniature: routing (this file), a script that talks back (scripts/context.mjs), a hook that fights back (scripts/hook.mjs), making sub-agents argue (reference/argue.md), and forcing divergence (reference/diverge.md). Use when the user types /spellbook <command>, or asks to argue, diverge, or set up the demo.license: Apache-2.0---# Spellbook (starter)The workshop build: a **pocket Impeccable** you assemble as the talk climbs. Itcritiques and re-imagines `sample/landing.html` (a deliberately-median page)using five dark arts in miniature. It runs as-is; every art has a `YOUR TURN`marker where you extend it. Read it top to bottom, then make it yours.## Setup (always do this first)DARK ART · *the script talks back.* Do not guess the project state. Run:```bashnode scripts/context.mjs```**Follow whatever it prints.** The script reads the environment (git branch,whether a NOTES.md exists) and returns a fresh, in-thread directive. Thatdirective outranks anything written statically below, because it is shaped tothe exact situation. If it prints `NO_NOTES`, create `NOTES.md` first. If itprints `READY`, continue.## Commands (the router)DARK ART · *route like a frontier model.* This file is a thin router. Do notinline the technique here; load the one expert reference the command needs, andnothing else, so context stays cheap.| Command | Load this, then follow it ||-----------------------------|---------------------------|| `/spellbook argue [page]` | `reference/argue.md` || `/spellbook diverge [page]` | `reference/diverge.md` |`[page]` defaults to `sample/landing.html`. If no command is given, ask which onethe user wants. Never run both references at once.## The hook (already working, even when this skill is not invoked)DARK ART · *hooks that fight back.* `scripts/hook.mjs` runs after every fileedit (wire it up via `.claude/settings.json`, see `README.md`). It scans thetouched file for a banned pattern and injects a system-reminder. You do notinvoke it; it watches. Try editing a file to include `TODO: fix later` andwatch it push back.

Frequently asked questions