Nguyen Anh Quan

You can write actionable structured notes with trace schemas and self-audit checklists, conduct pre-coding domain research with saved sources and indexed findings, reorganize and deduplicate shared notes directories, and autonomously create, test, benchmark, and optimize new skills from repeated work patterns.

They target engineers and agent operators running multi-agent or eval-driven environments who need durable shared memory. The create-notes skill specifically serves teams where future agents must act on prior experiment results, grader issues, build problems, and runtime discoveries recorded in a shared directory.

Trigger it whenever writing a file under notes/. It enforces one of four note variants, required frontmatter with a team-level creator filter, a structured-trace schema feeding a dashboard knowledge graph, a self-audit checklist, bundled stamp/lint/unattributed scripts, and shell-escaping guidance that prevents stripped markdown content.

Use it when you notice repeated tool sequences, recurring code patterns across attempts, or insights worth packaging as a reusable skill. It autonomously creates, tests, and optimizes skills from detected patterns, and also benchmarks and iterates on existing skills to improve their performance.

You need an environment with a shared notes directory structure (notes/, research/, experiments/) and an index.md file. The create-notes skill relies on its bundled scripts for stamping, linting, and attribution checks, and expects shell access where markdown escaping behavior must be handled carefully.