coretree-writing

Coretree Writing

Language: English | 한국어

Coretree is a compact multi-level bullet writing style for quick scanning.

It helps turn scattered notes, decisions, findings, and raw information into structured bullet trees that are easier to scan, compare, and revisit. By using shallow hierarchy, concise labels, and one-line relationships, Coretree makes decisions, causes, risks, and next actions easier to find without rereading long paragraphs.

It is intentionally small: not a template, not a knowledge base system, and not a universal writing method.

Why Coretree

Asking an agent to "use bullets to structure this document" does not always structure the information accurately. Coretree addresses this by clearly defining how to organize information with multi-level bullets. Coretree exists as a small named instruction for making bullet depth, importance, and promotion more predictable.

What It Is

Coretree uses multi-level bullets with explicit rules for depth, importance, and promotion.

• Main bullets cover core topics, decisions, or outcomes.
• Sub-bullets provide details, causes, conditions, evidence, examples, or follow-up information.
• Bullet depth shows dependency, not importance.
• Important details are promoted upward when they would otherwise be buried.
• Short labels such as Decision:, Cause:, Result:, and Status: can make scanning easier when the role of a bullet needs to be explicit.
• For short relationships, use : for label-value pairs and => for cause, change, or consequence.

When To Use It

Use Coretree when a reader needs to scan information quickly.

• Meeting notes become easier to revisit.
• Decision records show the decision, reason, and next action clearly.
• Research summaries separate findings from evidence.
• Bug reports show symptoms, causes, impact, and status without long paragraphs.
• AI agent output becomes more consistent and easier to review.

Quick Example

Before: project decision notes

The team needs to decide whether to keep the current Markdown-based documentation flow or move project notes into a database-backed knowledge base. The database option would make filtering and dashboards easier, but it would add setup work and make quick edits slower. The current Markdown flow is simple, works well with code review, and is easy for agents to edit. For now, the team decided to keep Markdown as the default and only revisit the knowledge base option if search and cross-project reporting become recurring problems.

After:

- Decision: keep Markdown as the default documentation flow
    - Simple to edit and review
    - Works well with code review
    - Easy for agents to modify

- Alternative: database-backed knowledge base
    - Benefit: easier filtering and dashboards
    - Cost: more setup work
    - Cost: slower quick edits

- Revisit condition
    - Search problems become recurring
    - Cross-project reporting becomes necessary

- Next action
    - Keep project notes in Markdown
    - Track search and reporting pain points

Where To Go Next

• Want more examples? Read EXAMPLES.md.
• Want the exact rules? Read SPEC.md.
• Want to use it as an agent skill? See Use As An Agent Skill.

Use As An Agent Skill

The GitHub project is named coretree-writing; the installable skill is named coretree.

The skill file is platform-neutral Markdown. Its compatibility field is kept as OpenCode metadata, but other agents can read the same instructions directly.

Install from the skill directory with the skills CLI. Use one or both commands depending on your agent:

# OpenCode
npx skills add https://github.com/netics01/coretree-writing/tree/main/skills/coretree -g -a opencode

# Claude Code
npx skills add https://github.com/netics01/coretree-writing/tree/main/skills/coretree -g -a claude-code

Manual paths:

• OpenCode: ~/.config/opencode/skills/coretree/SKILL.md
• Claude Code: ~/.claude/skills/coretree/SKILL.md
• Generic agents: copy skills/coretree/SKILL.md into the agent's skill directory, or provide it as Markdown instructions.

How To Ask An Agent To Use It

After installing the skill, ask your agent directly:

Write this document in Coretree style.

or:

Summarize this in Coretree writing.

To make Coretree the default style for a project, add an instruction to that project's agent guidance file, such as AGENTS.md:

When writing notes, summaries, decisions, findings, or documentation, use Coretree style unless the user asks for another format.

The exact trigger wording depends on the agent, but phrases such as Coretree style or Coretree writing are intended to load or apply the skill.

Language

• Primary language: English.
• Korean documents are companion versions for readers who prefer Korean.
• If English and Korean documents differ, use the English document.

Status

Coretree is intentionally small. The goal is not to define a broad writing framework, but to provide a clear convention that people and AI agents can apply consistently.

License

MIT License. See LICENSE.