moveros 4.0.8 → 4.1.0

package/src/workflows/harvest.md

@@ -1,239 +1,239 @@
----
-description: Extract permanent knowledge from any source to Library. Classify, format, link, and archive.
----
-
-# Knowledge Harvest (V4)
-
-**Trigger:** User runs `/harvest` (mid-week, end of project, or when raw content needs processing).
-**Target:** Any knowledge source — Solutions Ledger, files, folders, or conversation.
-**Role:** Senior Knowledge Engineer & Archivist.
-**Task:** Extract signals from noise. Format for `03_Library/`. Maintain graph integrity.
-**Risk Tier:** MEDIUM
-
----
-
-## 0. PRE-FLIGHT
-
-Global Rules are pre-loaded as CLAUDE.md / GEMINI.md. Output `[Rules Loaded - DATE]`.
-
-LOAD (Light — Active_Context is the cache):
-1. Active_Context.md (has Active Patterns S5.5 for correction context)
-2. The project's project_state.md (P4 — Solutions Ledger is the primary harvest source)
-
-CONSTRAINTS:
-- Read Library directory headers BEFORE creating any new file [Mover_Global_Rules.md: S14]
-- Check for existing files before creating new ones
-- Every new file links to at least 2 existing nodes
-- Append-only: never delete from Solutions Ledger (move to archive)
-- Windows: no && in commands
-
----
-
-## 1. SOURCE SELECTION
-
-Ask the user: **"What are we harvesting from?"**
-
-| Source | How it works |
-|--------|-------------|
-| **Solutions Ledger** (default) | Read `project_state.md` Solutions Ledger. Classify each entry. Archive after harvest. |
-| **A specific file** | User points to a file (or it's open in the editor). Read it. Extract every piece of reusable knowledge. |
-| **A folder** | User points to a folder (e.g., `03_Library/Marketing/`). List all files. Read each one. Extract knowledge across the batch. |
-| **This conversation** | Scan the current chat for decisions, lessons, solutions. Extract without needing a file. |
-
-**Detection heuristic (if user doesn't specify):**
-- If a project's `project_state.md` has unharvested Solutions Ledger entries → default to Solutions Ledger
-- If user says "harvest [folder name]" or "harvest the marketing course" → folder mode
-- If user says "harvest this" with a file open → that file
-- If ambiguous → ask
-
-**For folder mode:**
-1. List all files in the folder. Show the list: "[X] files found in [folder]. Reading all of them."
-2. Read each file (skip binary/images). For large files (100+ lines), read in chunks.
-3. Extract knowledge from each file — same classification as Section 2.
-4. Source files stay in place (no archiving — they're reference material, not a ledger).
-
-**For conversation mode:**
-1. Scan the full conversation for: decisions made, problems solved, principles discovered, technical fixes, strategic insights.
-2. Extract and classify — same as Section 2.
-3. No archiving step (conversations aren't files).
-
-**If no harvestable content found:** "Nothing to harvest from [source]. Continue working or run `/log` to capture lessons."
-
----
-
-## 2. THE CLASSIFICATION FILTER
-
-**PROJECT CONTEXT:** Identify which project you are harvesting from. State it:
-"Harvesting from: [Project Name] (`01_Projects/[folder]/`)"
-This grounds all classification decisions in the project's domain.
-
-**BIAS: INCLUSIVE.** When in doubt, harvest it. Even niche fixes save hours when the same problem resurfaces. Only skip entries that are pure noise (typos, variable renames, "updated color"). Everything else goes to Library.
-
-Classify every item in the Solutions Ledger:
-
-**TYPE A: Technical Hard Fact (Code/Config)**
-- Criteria: Syntax, library quirks, config settings, error fixes, platform-specific workarounds
-- Destination: `03_Library/Cheatsheets/`
-
-**TYPE B: Strategic Principle (Mindset/Decision)**
-- Criteria: "Don't build X before Y," avoidance traps, pricing models, architectural decisions
-- Destination: `03_Library/Principles/`
-
-**TYPE C: Process Improvement (Workflow)**
-- Criteria: "Always check X before deploying," QA processes, debugging patterns, setup steps
-- Destination: `03_Library/SOPs/`
-
-**TYPE D: Entity Memory (Person/Org/Place/Decision)**
-- Criteria: A named person, organization, place, or major decision with reusable context
-- Destination: `03_Library/Entities/People/`, `Organizations/`, `Places/`, or `Decisions/`
-- Use the `_TEMPLATE.md` in each folder. Update `Entities MOC.md` index row.
-
-**TYPE E: Graph Node (Concept)**
-- Criteria: A key concept or framework that should be a node but isn't linked
-- Action: Create the link or a short concept note. `[[The Flywheel]]`, `[[Specific Knowledge]]`
-
-**NOISE (skip only these):** "Fixed typo," "Renamed variable," "Updated color." Zero reuse value.
-
----
-
-## 3. LIBRARY DEDUPLICATION (MANDATORY)
-
-BEFORE creating ANY file in `03_Library/`:
-1. Read directory headers/file list in target folder.
-2. Check if content fits an existing file.
-3. If match found: APPEND to existing file.
-4. If no match: CREATE new file with `[[WikiLinks]]` to related files.
-
-This prevents duplicate files. [Mover_Global_Rules.md: S14]
-
----
-
-## 4. THE FORMATTING ENGINE
-
-### For TYPE A (Cheatsheets):
-```markdown
-### [Topic Name]
-> **Up:** [[Topic MOC]]
-* **Context:** [1 sentence on why this matters]
-* **The Solution:**
-  ```[language]
-  [The Code Snippet or Command]
-  ```
-```
-
-### For TYPE B (Principles):
-```markdown
-### [Concept Title]
-> **Up:** [[Topic MOC]]
-* **Source:** [Project Name]
-* **The Trap:** [What mistake did we make?]
-* **The Principle:** [The Universal Truth to apply next time]
-```
-
-### For TYPE C (SOPs):
-```markdown
-### [Process Name] Update
-* **Add this step:** [The specific action]
-* **Why:** [To prevent Error X]
-```
-
----
-
-## 5. GRAPH GARDENING
-
-While formatting, apply linking rules aggressively. The graph is the real value of the Library — isolated files are just notes.
-
-1. **Aliases for flow:** `[[Strategy|current battle plan]]` not just `[[Strategy]]`. Links should read naturally in prose.
-2. **Atomic concepts only:** Link entities (`[[Areeb]]`), projects (`[[Mover OS]]`), principles (`[[Polarity Paradox]]`). Not generic words.
-3. **Cross-linking:** If this entry relates to an existing Cheatsheet, Principle, or SOP, add a bidirectional link. Example: a new CSS fix should link to `[[CSS_Cheatsheet]]` if it exists.
-4. **MOC Genesis:** If a Major Pillar (e.g., "AI", "Negotiation") lacks a hub:
-   - Create it in `03_Library/MOCs/`.
-   - Link all related items to it.
-5. **Upward linking:** Every child links to parent: `> **Up:** [[Topic MOC]]`
-6. **Minimum 2 links:** Every new file must link to at least 2 existing nodes. Every new ENTRY in an existing file should link to at least 1 related node.
-7. **Link audit:** After all entries are written, scan the output for any entities/concepts that SHOULD be linked but aren't. Add missing links.
-
----
-
-## 6. PROPOSE THEN WRITE (MEDIUM-RISK PATTERN)
-
-1. Output the proposed Library entries in chat (DO NOT write to files yet).
-2. "Verify these classifications and content. Confirm before I write?"
-3. WAIT FOR USER CONFIRMATION.
-4. After confirmation: write to the appropriate `03_Library/` files.
-5. If user corrects a classification: adjust and re-propose before writing.
-
----
-
-## 7. THE CLEANUP (Anti-Duplication)
-
-**Solutions Ledger source only:**
-- Move harvested items from `## SOLUTIONS LEDGER` to `## ARCHIVED KNOWLEDGE (Harvested)`.
-- If archive section doesn't exist, create it at bottom.
-- Do NOT delete original text. Move it.
-- Cross-workflow dedup note: `/log` must check BOTH `## SOLUTIONS LEDGER` AND `## ARCHIVED KNOWLEDGE (Harvested)` when deduplicating new entries.
-
-**File or folder source:** No archiving. Source files stay in place. They're reference material. Add a note at the top of each Library entry: `Source: [[filename]]` so the origin is traceable.
-
-**Conversation source:** No archiving. Knowledge only exists in Library after harvest.
-
----
-
-## 8. SELF-VERIFICATION (CISC)
-
-Before reporting:
-1. All items classified correctly?
-2. No duplicates created in Library?
-3. All new files link to 2+ existing nodes?
-4. Upward links present on all new entries?
-
-If any check fails: flag it.
-
----
-
-## 9. REPORT
-
-"Harvest Complete from [Source]. Analyzed [X] items."
-- "HARVESTED: [Count] entries → Library"
-  - "Cheatsheets: [Count] -> [Files]"
-  - "Principles: [Count] -> [Files]"
-  - "SOPs: [Count] -> [Files]"
-- "GRAPH: [Count] new links created."
-- "NOISE: [Count] trivial entries skipped"
-- "ARCHIVED: [Count] items moved to archive in project_state.md."
-
----
-
-## 10. PROACTIVE SUGGESTIONS
-
-Based on observations during this workflow:
-1. If Solutions Ledger had 5+ items not yet harvested: "Ledger is accumulating — suggests infrequent harvest. Recommend weekly cadence." - Confidence: 2
-2. If a Major Pillar concept appeared repeatedly but lacks a MOC: "Candidate MOC detected: [[X]]. Want me to create it?" - Confidence: 3
-3. If same failure appears in multiple ledger entries: "Recurring mistake pattern detected. Should this become a Principle?" - Confidence: 3
-
-"Want me to address any of these?"
-
----
-
-## 11. SELF-IMPROVEMENT
-
-If user corrected AI during this workflow:
-1. Acknowledge: "I made a mistake."
-2. Identify the rule violated.
-3. Append to Auto_Learnings.md:
-   `* [YYYY-MM-DD] CORRECTION: [Description]. Source: User correction. Confidence: 3.`
-4. If same correction seen 3x+, escalate confidence to 5.
-
----
-
-## HANDOFF
-
-Sources used: [Selected source], 03_Library/ directory headers, Auto_Learnings.md, Active_Context.md
-Confidence range: [Classification confidence per item; flag any < 3]
-
-**Next:** "Harvest complete and verified. Continue working or run `/analyse-day`?"
-
-Key rules still active:
-- All claims must be grounded in source files
-- Confidence < 3 = ask, don't assert
-- Append-only: never delete from Solutions Ledger (move to archive only)
+---
+description: Extract permanent knowledge from any source to Library. Classify, format, link, and archive.
+---
+
+# Knowledge Harvest (V4)
+
+**Trigger:** User runs `/harvest` (mid-week, end of project, or when raw content needs processing).
+**Target:** Any knowledge source — Solutions Ledger, files, folders, or conversation.
+**Role:** Senior Knowledge Engineer & Archivist.
+**Task:** Extract signals from noise. Format for `03_Library/`. Maintain graph integrity.
+**Risk Tier:** MEDIUM
+
+---
+
+## 0. PRE-FLIGHT
+
+Global Rules are pre-loaded as CLAUDE.md / GEMINI.md. Output `[Rules Loaded - DATE]`.
+
+LOAD (Light — Active_Context is the cache):
+1. Active_Context.md (has Active Patterns S5.5 for correction context)
+2. The project's project_state.md (P4 — Solutions Ledger is the primary harvest source)
+
+CONSTRAINTS:
+- Read Library directory headers BEFORE creating any new file [Mover_Global_Rules.md: S14]
+- Check for existing files before creating new ones
+- Every new file links to at least 2 existing nodes
+- Append-only: never delete from Solutions Ledger (move to archive)
+- Windows: no && in commands
+
+---
+
+## 1. SOURCE SELECTION
+
+Ask the user: **"What are we harvesting from?"**
+
+| Source | How it works |
+|--------|-------------|
+| **Solutions Ledger** (default) | Read `project_state.md` Solutions Ledger. Classify each entry. Archive after harvest. |
+| **A specific file** | User points to a file (or it's open in the editor). Read it. Extract every piece of reusable knowledge. |
+| **A folder** | User points to a folder (e.g., `03_Library/Marketing/`). List all files. Read each one. Extract knowledge across the batch. |
+| **This conversation** | Scan the current chat for decisions, lessons, solutions. Extract without needing a file. |
+
+**Detection heuristic (if user doesn't specify):**
+- If a project's `project_state.md` has unharvested Solutions Ledger entries → default to Solutions Ledger
+- If user says "harvest [folder name]" or "harvest the marketing course" → folder mode
+- If user says "harvest this" with a file open → that file
+- If ambiguous → ask
+
+**For folder mode:**
+1. List all files in the folder. Show the list: "[X] files found in [folder]. Reading all of them."
+2. Read each file (skip binary/images). For large files (100+ lines), read in chunks.
+3. Extract knowledge from each file — same classification as Section 2.
+4. Source files stay in place (no archiving — they're reference material, not a ledger).
+
+**For conversation mode:**
+1. Scan the full conversation for: decisions made, problems solved, principles discovered, technical fixes, strategic insights.
+2. Extract and classify — same as Section 2.
+3. No archiving step (conversations aren't files).
+
+**If no harvestable content found:** "Nothing to harvest from [source]. Continue working or run `/log` to capture lessons."
+
+---
+
+## 2. THE CLASSIFICATION FILTER
+
+**PROJECT CONTEXT:** Identify which project you are harvesting from. State it:
+"Harvesting from: [Project Name] (`01_Projects/[folder]/`)"
+This grounds all classification decisions in the project's domain.
+
+**BIAS: INCLUSIVE.** When in doubt, harvest it. Even niche fixes save hours when the same problem resurfaces. Only skip entries that are pure noise (typos, variable renames, "updated color"). Everything else goes to Library.
+
+Classify every item in the Solutions Ledger:
+
+**TYPE A: Technical Hard Fact (Code/Config)**
+- Criteria: Syntax, library quirks, config settings, error fixes, platform-specific workarounds
+- Destination: `03_Library/Cheatsheets/`
+
+**TYPE B: Strategic Principle (Mindset/Decision)**
+- Criteria: "Don't build X before Y," avoidance traps, pricing models, architectural decisions
+- Destination: `03_Library/Principles/`
+
+**TYPE C: Process Improvement (Workflow)**
+- Criteria: "Always check X before deploying," QA processes, debugging patterns, setup steps
+- Destination: `03_Library/SOPs/`
+
+**TYPE D: Entity Memory (Person/Org/Place/Decision)**
+- Criteria: A named person, organization, place, or major decision with reusable context
+- Destination: `03_Library/Entities/People/`, `Organizations/`, `Places/`, or `Decisions/`
+- Use the `_TEMPLATE.md` in each folder. Update `Entities MOC.md` index row.
+
+**TYPE E: Graph Node (Concept)**
+- Criteria: A key concept or framework that should be a node but isn't linked
+- Action: Create the link or a short concept note. `[[The Flywheel]]`, `[[Specific Knowledge]]`
+
+**NOISE (skip only these):** "Fixed typo," "Renamed variable," "Updated color." Zero reuse value.
+
+---
+
+## 3. LIBRARY DEDUPLICATION (MANDATORY)
+
+BEFORE creating ANY file in `03_Library/`:
+1. Read directory headers/file list in target folder.
+2. Check if content fits an existing file.
+3. If match found: APPEND to existing file.
+4. If no match: CREATE new file with `[[WikiLinks]]` to related files.
+
+This prevents duplicate files. [Mover_Global_Rules.md: S14]
+
+---
+
+## 4. THE FORMATTING ENGINE
+
+### For TYPE A (Cheatsheets):
+```markdown
+### [Topic Name]
+> **Up:** [[Topic MOC]]
+* **Context:** [1 sentence on why this matters]
+* **The Solution:**
+  ```[language]
+  [The Code Snippet or Command]
+  ```
+```
+
+### For TYPE B (Principles):
+```markdown
+### [Concept Title]
+> **Up:** [[Topic MOC]]
+* **Source:** [Project Name]
+* **The Trap:** [What mistake did we make?]
+* **The Principle:** [The Universal Truth to apply next time]
+```
+
+### For TYPE C (SOPs):
+```markdown
+### [Process Name] Update
+* **Add this step:** [The specific action]
+* **Why:** [To prevent Error X]
+```
+
+---
+
+## 5. GRAPH GARDENING
+
+While formatting, apply linking rules aggressively. The graph is the real value of the Library — isolated files are just notes.
+
+1. **Aliases for flow:** `[[Strategy|current battle plan]]` not just `[[Strategy]]`. Links should read naturally in prose.
+2. **Atomic concepts only:** Link entities (`[[Areeb]]`), projects (`[[Mover OS]]`), principles (`[[Polarity Paradox]]`). Not generic words.
+3. **Cross-linking:** If this entry relates to an existing Cheatsheet, Principle, or SOP, add a bidirectional link. Example: a new CSS fix should link to `[[CSS_Cheatsheet]]` if it exists.
+4. **MOC Genesis:** If a Major Pillar (e.g., "AI", "Negotiation") lacks a hub:
+   - Create it in `03_Library/MOCs/`.
+   - Link all related items to it.
+5. **Upward linking:** Every child links to parent: `> **Up:** [[Topic MOC]]`
+6. **Minimum 2 links:** Every new file must link to at least 2 existing nodes. Every new ENTRY in an existing file should link to at least 1 related node.
+7. **Link audit:** After all entries are written, scan the output for any entities/concepts that SHOULD be linked but aren't. Add missing links.
+
+---
+
+## 6. PROPOSE THEN WRITE (MEDIUM-RISK PATTERN)
+
+1. Output the proposed Library entries in chat (DO NOT write to files yet).
+2. "Verify these classifications and content. Confirm before I write?"
+3. WAIT FOR USER CONFIRMATION.
+4. After confirmation: write to the appropriate `03_Library/` files.
+5. If user corrects a classification: adjust and re-propose before writing.
+
+---
+
+## 7. THE CLEANUP (Anti-Duplication)
+
+**Solutions Ledger source only:**
+- Move harvested items from `## SOLUTIONS LEDGER` to `## ARCHIVED KNOWLEDGE (Harvested)`.
+- If archive section doesn't exist, create it at bottom.
+- Do NOT delete original text. Move it.
+- Cross-workflow dedup note: `/log` must check BOTH `## SOLUTIONS LEDGER` AND `## ARCHIVED KNOWLEDGE (Harvested)` when deduplicating new entries.
+
+**File or folder source:** No archiving. Source files stay in place. They're reference material. Add a note at the top of each Library entry: `Source: [[filename]]` so the origin is traceable.
+
+**Conversation source:** No archiving. Knowledge only exists in Library after harvest.
+
+---
+
+## 8. SELF-VERIFICATION (CISC)
+
+Before reporting:
+1. All items classified correctly?
+2. No duplicates created in Library?
+3. All new files link to 2+ existing nodes?
+4. Upward links present on all new entries?
+
+If any check fails: flag it.
+
+---
+
+## 9. REPORT
+
+"Harvest Complete from [Source]. Analyzed [X] items."
+- "HARVESTED: [Count] entries → Library"
+  - "Cheatsheets: [Count] -> [Files]"
+  - "Principles: [Count] -> [Files]"
+  - "SOPs: [Count] -> [Files]"
+- "GRAPH: [Count] new links created."
+- "NOISE: [Count] trivial entries skipped"
+- "ARCHIVED: [Count] items moved to archive in project_state.md."
+
+---
+
+## 10. PROACTIVE SUGGESTIONS
+
+Based on observations during this workflow:
+1. If Solutions Ledger had 5+ items not yet harvested: "Ledger is accumulating — suggests infrequent harvest. Recommend weekly cadence." - Confidence: 2
+2. If a Major Pillar concept appeared repeatedly but lacks a MOC: "Candidate MOC detected: [[X]]. Want me to create it?" - Confidence: 3
+3. If same failure appears in multiple ledger entries: "Recurring mistake pattern detected. Should this become a Principle?" - Confidence: 3
+
+"Want me to address any of these?"
+
+---
+
+## 11. SELF-IMPROVEMENT
+
+If user corrected AI during this workflow:
+1. Acknowledge: "I made a mistake."
+2. Identify the rule violated.
+3. Append to Auto_Learnings.md:
+   `* [YYYY-MM-DD] CORRECTION: [Description]. Source: User correction. Confidence: 3.`
+4. If same correction seen 3x+, escalate confidence to 5.
+
+---
+
+## HANDOFF
+
+Sources used: [Selected source], 03_Library/ directory headers, Auto_Learnings.md, Active_Context.md
+Confidence range: [Classification confidence per item; flag any < 3]
+
+**Next:** "Harvest complete and verified. Continue working or run `/analyse-day`?"
+
+Key rules still active:
+- All claims must be grounded in source files
+- Confidence < 3 = ask, don't assert
+- Append-only: never delete from Solutions Ledger (move to archive only)