Learn

Learn

This skill wraps the preserved .learn/ workflow as a standard Agent Skill.

Use this skill when the user wants to:

• generate a learning plan (learn-plan)
• write a Chinese ML systems learning article (learn-write)
• review or optionally translate an article (learn-review)
• add a published article to the knowledge graph (learn-add)

Canonical Resources

The .learn/ directory is intentionally preserved. Do not move it while using this skill.

Load these files as needed:

• .learn/skill.md: full workflow definition and command semantics
• .learn/config.md: agent preferences and adjustable behavior
• .learn/index/style-guide.md: writing style and quality guidance
• .learn/templates/: article templates for paper reading, tutorials, system design, and code walkthroughs
• .learn/example/knowledge-graph.json: published article metadata and relationships, but do not use this, as this is the original author's knowledge graph and cannot apply directly to you. Instead, please use it as a reference for how to structure your own knowledge graph.

Workflow

1. Read .learn/skill.md first to identify the relevant subcommand workflow.
2. Route the task by intent:
   • planning work follows the learn-plan flow
   • writing work follows the learn-write flow
   • review or translation work follows the learn-review flow
   • knowledge graph updates follow the learn-add flow
3. Load only the supporting .learn/ files required by that flow.
4. Preserve the hard constraints from .learn/skill.md, especially Chinese-first writing, excluding [Pending Review] articles as references, and requiring commit hashes for external source-code citations.

Compatibility Notes

Claude Code can discover this skill through .claude/skills/learn/SKILL.md, which is only a wrapper. OpenAI local shell usage should pass this directory, .agents/skills/learn, as the skill path.