adderall
v1.3.0
Published
Dosage-based meta-skill pack for Claude, Cursor, Codex, and Hermes Agent. Control how strictly an agent follows another skill — not just what it does.
Downloads
121
Maintainers
Readme
Install in one line
npx adderall install allThat's it. No clone, no build step, no Python, no Docker. The CLI ships the seven SKILL.md files inside the npm package and copies them to the correct location for each platform.
Prefer to pick platforms individually:
npx adderall install claude # → ~/.claude/skills/adderall-<dose>
npx adderall install cursor # → ~/.cursor/skills/adderall-<dose>
npx adderall install codex # → ~/.codex/skills/ + AGENTS.md block
npx adderall install hermes # → ~/.hermes/skills/adderall-<dose>Pin adderall to the current repo instead of the user profile:
npx adderall install cursor --project # → ./.cursor/skills/adderall-<dose>
npx adderall install codex --project # → ./.codex/skills/ + ./AGENTS.md blockRemove cleanly:
npx adderall uninstall allAudit what's installed and where:
npx adderall doctorSee npx adderall help for every command. Detailed per-platform rules live in INSTALL.md.
RX Shelf
Dosage Guide
| Skill | Adherence | Flexibility | Intended Use |
| ------------------ | :-------: | :---------: | ------------------------- |
| adderall-5mg | 0.10 | 0.90 | Open-ended exploration |
| adderall-7.5mg | 0.25 | 0.75 | Flexible guidance |
| adderall-10mg | 0.50 | 0.50 | Balanced execution |
| adderall-12.5mg | 0.70 | 0.30 | High adherence |
| adderall-15mg | 0.85 | 0.15 | Near-strict execution |
| adderall-20mg | 0.95 | 0.05 | Strict execution |
| adderall-30mg | 1.00 | 0.00 | Maximal literal adherence |
How It Works
Prefix a target skill invocation with a dosage skill:
/adderall-10mg /frontend-ui build a login page
/adderall-20mg /python-debug investigate failing tests
/adderall-5mg /research-arxiv survey recent work on sparse autoencodersAt runtime:
- The dosage skill resolves the target skill that follows it.
- It sets the expected adherence (how literally instructions must be followed) and flexibility (how much initiative the agent may take).
- The agent executes the target skill through that behavioral lens — no edits to the target skill required.
One SKILL.md, Four Platforms
A single SKILL.md per dosage is valid on every supported platform. No per-platform forks; no duplicate content. Each platform reads the fields it understands and ignores the rest.
| Platform | Skills path | Activation |
| ------------------- | ----------------------------------------- | ----------------------------------------------------------- |
| Claude Code | ~/.claude/skills/adderall-<dose>/ | Auto-loaded; matches description triggers |
| Claude Desktop | Upload a .zip per skill via the UI | Toggle in the skill panel |
| Cursor | ~/.cursor/skills/adderall-<dose>/ | Listed in the skill picker / slash menu |
| Codex CLI | ~/.codex/skills/adderall-<dose>/ + AGENTS.md block | Auto-loaded; /adderall-<dose> triggers execution |
| Hermes Agent | ~/.hermes/skills/adderall-<dose>/ | hermes skills list |
Choosing a Dosage
- Low dosages (
5mg,7.5mg) — prefer when you want optional ideas, broader exploration, or creative extensions beyond the target skill. - Mid dosage (
10mg) — prefer when the target skill should be respected but not over-applied. - High dosages (
12.5mg→30mg) — prefer when consistency, repeatability, and literal compliance matter more than initiative.
CLI Reference
npx adderall <command> [options]
Commands:
install <platform> Install adderall on a platform (or 'all')
uninstall <platform> Remove adderall from a platform (or 'all')
doctor Report where adderall is installed
list List the 7 dosages and their profiles
info <dose> Print a dosage's SKILL.md
help Show the full help screen
Options:
--project Install at project scope (./.<platform>/skills)
--link Use symlinks instead of copies (dev mode)
Platforms:
claude, cursor, codex, hermes, allSKILL.md Format
Every skill uses a single unified frontmatter that is valid on all four target platforms:
---
name: adderall-10mg
description: Balanced execution dosage — adherence 0.50, flexibility 0.50. Use when the user prefixes a target skill with /adderall-10mg.
version: 1.0.0
author: adhdcreator
license: MIT
metadata:
hermes:
tags: [Meta, Control, Dosage, adderall]
related_skills: [adderall-7.5mg, adderall-12.5mg]
---- Claude / Cursor read the top-level
nameanddescriptionand treat the body as instructions. Extra fields are ignored. - Codex reads the description via the
AGENTS.mdblock; skill bodies are accessible under~/.codex/skills/. - Hermes reads the full frontmatter including
metadata.hermes.tagsandrelated_skills.
See templates/SKILL.template.md for the scaffold, AUTHORING.md for conventions, and INSTALL.md for the exact rules each platform enforces.
Repository Structure
adderall/
├── bin/
│ └── adderall.js # CLI entry point (published as `adderall` bin)
├── src/
│ ├── cli.js # argv parser + command dispatcher
│ ├── logger.js # zero-dep ANSI logger
│ ├── fsutil.js # zero-dep fs helpers (copy / symlink / rm)
│ └── platforms.js # claude + cursor + codex + hermes adapters
├── skills/
│ ├── adderall-5mg/SKILL.md
│ ├── adderall-7.5mg/SKILL.md
│ ├── adderall-10mg/SKILL.md
│ ├── adderall-12.5mg/SKILL.md
│ ├── adderall-15mg/SKILL.md
│ ├── adderall-20mg/SKILL.md
│ └── adderall-30mg/SKILL.md
├── templates/
│ └── SKILL.template.md
├── assets/
│ ├── banner.png # Top banner (pharmacy-grade cover art)
│ ├── bannerpills.png # RX shelf — one bottle per dosage
│ └── {banner,pills}.{txt,ans} # CP437 ASCII fallbacks
├── AUTHORING.md # How skills in this repo are authored
├── CHANGELOG.md
├── INSTALL.md # Detailed per-platform install guide
├── LICENSE # MIT, (c) adhdcreator
├── package.json # npm manifest (bin: "adderall")
└── README.mdAll skills in this repository are authored by adhdcreator. External contributions are not accepted; issues and discussions are welcome.
Philosophy
A skill tells the agent what to do. A dosage tells the agent how much of itself to bring to it.
adderall treats adherence as a first-class, tunable parameter — the same way temperature tunes sampling. This keeps target skills pure (no scattered "be strict" / "be creative" flags) and makes agent behavior reproducible across sessions, users, and platforms.
License
MIT (c) adhdcreator. See LICENSE.
