MindGraph 🕸️🧠

A knowledge graph + learnable process engine for OpenClaw. Two core features:

1. 1. Knowledge Graph — Obsidian-style [[wikilinks]] across all workspace files
2. MindSkills — Learned, repeatable processes that save graph-connected results

Always-On Rules

Every time you write or edit a markdown file, use [[wikilinks]] for:

• - People: [[Alice]], INLINECODE3
• Projects: [[my-saas]], INLINECODE5
• Companies/tools: [[Stripe]], [[Vercel]], INLINECODE8
• Concepts/frameworks: [[Knockout Test]], INLINECODE10
• Other agents/models: [[Claude Code]], INLINECODE12

This is not optional. Links are how knowledge connects. No links = isolated notes = useless.

NEVER use [[wikilinks]] in messages to the user (Telegram, Discord, etc.). Wikilinks are for workspace files only. In conversations, write names plain: "Alice", not "[[Alice]]".

After significant file changes, rebuild the index:
CODEBLOCK0

Graph Commands

CODEBLOCK1

MindSkills — Learned Processes

MindSkills are repeatable frameworks stored in skills/mindgraph/mindskills/. Each has a defined process and saves results as graph-connected markdown.

Using a MindSkill

CODEBLOCK2

When a user asks to run a process (e.g., "run the knockout test on X"), follow this flow:

1. 1. Read the mindskill's PROCESS.md for the process definition
2. Execute the process conversationally
3. Save the result to INLINECODE16
4. Use [[wikilinks]] throughout the result file
5. Include YAML frontmatter with metadata
6. Rebuild the graph index

Result file template:
CODEBLOCK3

Learning a New MindSkill

When a user says "learn a mindskill called X" or describes a repeatable process:

CODEBLOCK4

This creates the directory structure. Then write the PROCESS.md based on the user's description.

A good PROCESS.md contains:

• - Purpose: What this process does and when to use it
• Trigger phrases: What the user might say to invoke this
• Steps: The actual process to follow (numbered)
• Output format: What the result file should contain
• Verdict/scoring: How to summarize the outcome (if applicable)

Discovering MindSkills

When a user's request matches a learned mindskill, proactively suggest it:

• - "Want me to run the [[Knockout Test]] on that?"
• "I have an [[SEO Validator]] mindskill — should I audit that?"
• "This looks like a [[Competitor Analysis]] — want the full framework?"

Link Resolution

Links match (case-insensitive) against:

1. 1. File basenames: [[MEMORY]] → INLINECODE19
2. Project dirs: [[my-saas]] → INLINECODE21
3. MindSkill results: [[Pet Tracker KT]] → knockout test result
4. YAML aliases: aliases: [AV-Check] → [[AV-Check]] resolves
5. Unresolved → concept node (still tracked for backlinks)

File Locations

• - Graph index: mindgraph.json (workspace root)
• MindSkills: INLINECODE26
• Script: INLINECODE27