napkin

Read and curate the Napkin Runbook format at session start.

1|Updated Mar 27, 2026

Search Tags:#automation #curation #agent #memory #runbook #knowledge-base #napkin

Ready to run

npx skills add https://github.com/NEXUZ-SYS/devflow --skill napkin-nexuz-sys

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: napkin
Source: https://github.com/NEXUZ-SYS/devflow/tree/main/skills/napkin
Command: npx skills add https://github.com/NEXUZ-SYS/devflow --skill napkin-nexuz-sys

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Keeps a continuously curated runbook of recurring mistakes, corrections, and patterns to guide agents, reducing repeated errors and accelerating reliable execution.

Core Features & Use Cases

Always-active memory: reads and applies the napkin guidance at session start.
Structured templates: provides a standard format, categories, and agent sections for consistent notes.
Curation policy: on every read, re-prioritize, merge duplicates, remove stale items, and enforce caps.
Agent-specific notes: supports dedicated subsections for domain-specific learnings.
Mode compatibility: works across Full, Lite, and Minimal operation modes.

Quick Start

Start a session by loading the napkin runbook and consult it before taking action.

Frequently Asked Questions about napkin

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I stop my AI agent from repeating the same mistakes across sessions?▼

A persistent memory runbook prevents repeated mistakes by logging corrections and patterns. The agent reads this guidance at session start to apply learned guardrails, ensuring reliable execution across projects.

What is the best way to maintain an agent knowledge-base without accumulating stale data?▼

Maintaining an agent knowledge-base requires an automated curation policy that re-prioritizes items, merges duplicates, removes stale entries, and enforces category limits on every read, keeping the runbook continuously optimized.

How do I structure agent memory to support domain-specific learnings for different tasks?▼

Structure agent memory with dedicated per-agent sections within a standard template. This format separates domain-specific notes and categorized learnings to guide individual agents while maintaining a unified runbook structure.

Can I use a markdown runbook to enforce guardrails for automated agents?▼

Yes, you can enforce guardrails using a markdown runbook at .context/napkin.md. The agent reads this file at session start to apply established patterns and restrictions before taking any project actions.

Does the agent memory curation work across different operation modes?▼

Agent memory curation works across Full, Lite, and Minimal operation modes. It consistently reads the standard template and applies the curation policy to maintain the runbook regardless of the selected mode.

Why do I need a curation policy for my agent's persistent memory?▼

A curation policy for persistent memory prevents unbounded growth and redundancy. It automatically enforces caps, merges duplicates, and removes stale items on every read, keeping the runbook actionable and concise.