skill

nanoForge

A lightweight, universal AI coding methodology in a single SKILL.md. Covers task sizing (XS→XL), strategy selection, code analysis, user interaction, and execution discipline — distilled from Superpow

TreeberryTomato/nanoForgecommunitylistedunreviewed

Use skill

Set up

Skills are installed by placing their folder where your agent loads skills.

Download ZIP from GitHub Open the repository

The download comes from github.com/TreeberryTomato/nanoForge itself, not from our server, so what you get is exactly what its author published.

Install instructions come from the README

Skills do not use an MCP config. Follow the steps in the README below.

README

synced Oct 23, 2026

NanoForge

A lightweight, universal AI coding methodology in a single SKILL.md.

Nano = minimal, precise, zero bloat. Forge = craft code with discipline.

NanoForge distills the essence of Superpowers, Karpathy Guidelines, Matt Pocock, GStack, and Addy Osmani's agent skills into one composable file — ~130 lines, zero dependencies, works with any AI coding agent.

Why NanoForge?

vs	Superpowers	Karpathy	Matt Pocock	NanoForge
Files	14 SKILL.md	1 SKILL.md	17+ SKILL.md	1 SKILL.md
Lines	~2000+	~65	~2000+	~130
Task sizing (XS→XL)	✓	✗	Partial	✓
Strategy selection	✓ (TDD/subagent)	✗	✓ (tracer bullets)	✓ 7 modes
Code analysis	✓ (debugging)	✗	✗	✓ trace path
User interaction	✓ (brainstorming)	✗	✓ (grill)	✓ protocol
Execution discipline	✓ (verification)	✓ (surgical)	✓	✓ 5 gates

Install

Copy the file

# Claude Code
cp SKILL.md ~/.claude/skills/nanoForge/SKILL.md

# OpenCode
mkdir -p .opencode/skills/nanoForge
cp SKILL.md .opencode/skills/nanoForge/SKILL.md

# Cursor
cp SKILL.md .cursor/skills/nanoForge/SKILL.md

# Codex CLI
cp SKILL.md ~/.codex/skills/nanoForge/SKILL.md

# Gemini CLI
cp SKILL.md ~/.gemini/skills/nanoForge/SKILL.md

Via npm

npx nano-forge

How It Works

Task comes in
  → Size it (XS/S/M/L/XL + risk modifier)
  → Read relevant code (trace path)
  → Choose strategy (Direct / Plan→Execute / TDD / Incremental / Subagent)
  → Interact (ask questions → propose options → get sign-off)
  → Execute with discipline (5 verification gates)
  → Report done

1. Task Assessment

Size	Scope	Effort	Approach
XS	1 file, single change	<1 min	Direct edit
S	1-2 files	1-5 min	Brief plan
M	3-5 files	5-15 min	Plan + TDD
L	5-8 files	15-60 min	Formal plan
XL	8+ files	60+ min	Break down

2. Code Analysis

Trace the request → entry point → data flow → impact scope → side effects.

3. Strategy Selection

7 modes: Direct, Plan→Execute, TDD, Incremental slices, Subagent, Doubt-driven, Investigate→Fix→Verify.

4. User Interaction

One question at a time. Present options with trade-offs. Status vocabulary: DONE / DONE_WITH_CONCERNS / BLOCKED / NEEDS_CLARIFICATION.

5. Execution Discipline

5 verification gates for every change: build, tests, lint, behavior, scope.

Principles

Surgical changes — change only what the task requires
Simplicity first — simplest design that works
Evidence over claims — verify before declaring done
Fail fast — surface blockers early

Compatible Agents

Claude Code · OpenCode · Cursor · Codex CLI · Gemini CLI · GitHub Copilot CLI · Kiro · Windsurf · Antigravity

License

MIT