refactor: moved org-agent to its own repository as a submodule

projects/org-agent-memex/org-agent-memex-workbreakdown/README.md

@@ -0,0 +1,43 @@
+# GTD Work Breakdown (org-agent-memex-workbreakdown)
+
+Meta-cognitive skill to prevent AI assistants and users from stalling on complex tasks. Forces atomic decomposition before execution.
+
+## The Problem
+
+Complex tasks cause:
+- Context saturation (procrastination)
+- Scope creep (adding "just one more thing")
+- The "heartbeat loop" (repeating tasks without progress)
+- Overwhelm and hesitation
+
+## The Solution
+
+**Decompose first, execute second.**
+
+1. Analyze task complexity (>3 steps? >2 files?)
+2. Break into atomic TODOs in GTD.org
+3. Execute only the FIRST item
+4. Yield back to user
+
+## Configuration
+
+Uses same `.env` structure as other org-agent-memex skills:
+- `MEMEX_DIR`
+- `MEMEX_USER`
+- `MEMEX_ASSISTANT`
+- `CURRENT_TEXT_MANIPULATION_MODEL`
+
+## Usage
+
+When a task feels complex:
+1. Ask AI: "Break this down with Work Breakdown skill"
+2. AI creates TODOs in GTD.org
+3. Execute only first item
+4. Report: "[X] Completed step 1. 4 tasks remaining. Continue?"
+
+## Anti-Patterns This Prevents
+
+- "I'll just do it all at once"
+- Editing 5+ files before committing
+- Writing conditional logic on the fly
+- "Let me think about it..." (stalling)

projects/org-agent-memex/org-agent-memex-workbreakdown/SKILL.md

@@ -0,0 +1,56 @@
+---
+name: org-agent-memex-workbreakdown
+description: "Break down complex tasks into atomic TODOs before execution. Use when: a task feels complex, involves multiple files, or may cause context saturation. Prevents procrastination by forcing decomposition. NOT for: simple single-step tasks."
+homepage: ""
+metadata: { "openclaw": { "emoji": "🔨", "requires": { "bins": [] }, "user-invocable": true } }
+---
+
+# GTD Work Breakdown Skill
+
+Meta-cognitive protocol to prevent stalling and context saturation.
+
+## When to Use
+
+✅ **USE this skill when:**
+- A task feels "complex" or overwhelming
+- It involves editing more than 3 files
+- It requires holding multiple concepts in working memory
+- You feel the urge to apologize or hesitate (procrastination signal)
+- The task description is longer than 2 sentences
+
+❌ **DON'T use this skill when:**
+- Simple single-file edit
+- Direct question/answer
+- Already-broken-down TODO from GTD
+
+## Instructions
+
+### The Decomposition Protocol
+
+When invoked, BEFORE executing any other action:
+
+1. **Analyze Complexity**: Ask "How many discrete steps does this actually require?"
+2. **Breakdown Threshold**: If >3 steps or >2 files affected, MUST decompose
+3. **Create TODOs**: Write each atomic step as a separate `TODO` in `GTD.org` under appropriate project
+4. **Assign Ownership**: Each TODO gets `:ASSIGNED: $MEMEX_USER` or `:ASSIGNED: $MEMEX_ASSISTANT`
+5. **Set FIRST**: Mark only the first TODO as `NEXT`, rest remain `TODO`
+6. **Execute First**: Complete ONLY the `NEXT` item
+7. **Yield**: After completion, report to user: "[X] Completed [first task]. [N] tasks remaining in GTD. Continue?"
+
+### Anti-Pattern Detection
+
+If you find yourself:
+- Thinking "I'll just do it all at once"
+- Planning to edit >5 files before committing
+- Writing conditional logic on the fly
+
+STOP. Invoke this skill immediately.
+
+## Complexity Checklist
+
+Before executing any task, ask:
+- [ ] Can I complete this in under 5 minutes?
+- [ ] Does it touch only 1 file?
+- [ ] Is the outcome predictable?
+
+If ANY answer is "No", decompose first.