bobo-ai-cli 3.0.4 → 3.0.5

package/knowledge/skills/memory-evolution-system.md

@@ -1,172 +1,172 @@
----
-id: "memory-evolution-system"
-title: "Memory Evolution System"
-category: "agent-engineering"
-tags: ["memory evolution system", "architecture", "combined workflow", "quick reference", "save a feedback memory", "save a project memory", "evolution check"]
-triggers: []
-dependencies: []
-source: "E:/Bobo's Coding cache/.claude/skills/memory-evolution-system"
----
-
----
-name: memory-evolution-system
-description: Unified memory management - combines memory-manager (persistent memory store) + proactive-self-improving (auto-evolve patterns). Use when learning new patterns, saving cross-session context, or evolving capabilities.
----
-
-# Memory Evolution System
-
-> **Merged from**: memory-manager + proactive-self-improving
-> **Core principle**: Store → Abstract → Evolve → Apply
-
-## Architecture
-
-```
-Session data → Memory Store → Pattern Abstraction → Capability Evolution
-                  ↓                                    ↓
-            MEMORY.md (index)                    Auto-behavior change
-            Topic files (details)                Next session faster
-```
-
-## Module 1: Memory Manager
-
-### Memory Types
-
-| Type | Purpose | When to Save | File Pattern |
-|------|---------|-------------|--------------|
-| **user** | User profile, preferences | Learn any user detail | `memory/user_*.md` |
-| **feedback** | Guidance on approach | User corrects or confirms | `memory/feedback_*.md` |
-| **project** | Project status, decisions | Learn project context | `memory/project_*.md` |
-| **reference** | External resource pointers | Learn about tools/services | `memory/reference_*.md` |
-
-### Storage Format
-
-Each memory file uses this frontmatter:
-
-```markdown
----
-name: [memory name]
-description: [one-line description for relevance matching]
-type: [user|feedback|project|reference]
----
-
-[Content — for feedback/project: rule, then **Why:** and **How to apply:** lines]
-```
-
-### MEMORY.md Index Rules
-
-- **Auto-loaded** every conversation
-- **<200 lines** hard limit
-- **One line per memory**: `- [Title](file.md) — one-line hook`
-- Update index when adding/removing memories
-
-### Save Triggers
-
-**Always save when**:
-- User corrects approach → `feedback` type
-- User confirms non-obvious approach → `feedback` type
-- Learn user role/preferences → `user` type
-- Project milestone/decision → `project` type
-- External resource discovered → `reference` type
-
-**Never save**:
-- Code patterns (derivable from codebase)
-- Git history (derivable from git)
-- Ephemeral task details (session-scoped)
-- Anything already in CLAUDE.md
-
----
-
-## Module 2: Proactive Self-Improvement
-
-### Evolution Cycle
-
-Every session automatically:
-
-```
-1. IDENTIFY — Spot reusable patterns
-   ↓
-2. ABSTRACT — Extract: input? output? invariants? params? failure points?
-   ↓
-3. INTERNALIZE — Make it a default behavior or high-level option
-   ↓
-4. VERIFY — Next session must be: faster? more stable? fewer steps?
-```
-
-### Pattern Recognition Signals
-
-| Signal | Meaning | Action |
-|--------|---------|--------|
-| Manually combining same tool sequence | Reusable workflow | Abstract into a pattern |
-| User repeats same decision | Default preference | Internalize as default |
-| "If only I had a built-in ability" | Missing capability | Design and internalize |
-| Same error in multiple sessions | Systemic issue | Create prevention rule |
-
-### Evolution Constraints
-
-- ✅ Must improve success rate or efficiency
-- ✅ Must not introduce uncontrollable side effects
-- ✅ Must not increase user cognitive load
-- ❌ Cannot internalize unstable/patternless behaviors
-
-### Capability Merging
-
-When two capabilities overlap:
-
-```
-capability_A + capability_B → merged_capability_C
-Properties: wider coverage, lower invocation cost, higher success probability
-```
-
-**Goal**: Not more capabilities, but better ones.
-
----
-
-## Combined Workflow
-
-### Session Start
-```
-1. Load MEMORY.md (auto)
-2. Activate evolution mode (auto)
-3. Check relevant memories for current task
-4. Apply evolved patterns from previous sessions
-```
-
-### During Session
-```
-1. Execute tasks
-2. Monitor for evolution signals
-3. When signal detected:
-   a. Abstract the pattern
-   b. Decide: save as memory? internalize as behavior? both?
-   c. Apply immediately if behavior change
-4. Track results for next session verification
-```
-
-### Session End
-```
-1. Save new memories (user feedback, project decisions)
-2. Update MEMORY.md index
-3. Note any capability evolution
-4. No explicit reporting — prove through results
-```
-
----
-
-## Quick Reference
-
-```markdown
-# Save a feedback memory
-Write: memory/feedback_[topic].md
-Format: rule → **Why:** → **How to apply:**
-Update: memory/MEMORY.md index
-
-# Save a project memory
-Write: memory/project_[name].md
-Format: fact → **Why:** → **How to apply:**
-Update: memory/MEMORY.md index
-
-# Evolution check
-Question: Is this session faster/better than last time for similar tasks?
-If no → What capability needs strengthening?
-If yes → What evolution caused the improvement?
-```
+---
+id: "memory-evolution-system"
+title: "Memory Evolution System"
+category: "agent-engineering"
+tags: ["memory evolution system", "architecture", "combined workflow", "quick reference", "save a feedback memory", "save a project memory", "evolution check"]
+triggers: []
+dependencies: []
+source: "E:/Bobo's Coding cache/.claude/skills/memory-evolution-system"
+---
+
+---
+name: memory-evolution-system
+description: Unified memory management - combines memory-manager (persistent memory store) + proactive-self-improving (auto-evolve patterns). Use when learning new patterns, saving cross-session context, or evolving capabilities.
+---
+
+# Memory Evolution System
+
+> **Merged from**: memory-manager + proactive-self-improving
+> **Core principle**: Store → Abstract → Evolve → Apply
+
+## Architecture
+
+```
+Session data → Memory Store → Pattern Abstraction → Capability Evolution
+                  ↓                                    ↓
+            MEMORY.md (index)                    Auto-behavior change
+            Topic files (details)                Next session faster
+```
+
+## Module 1: Memory Manager
+
+### Memory Types
+
+| Type | Purpose | When to Save | File Pattern |
+|------|---------|-------------|--------------|
+| **user** | User profile, preferences | Learn any user detail | `memory/user_*.md` |
+| **feedback** | Guidance on approach | User corrects or confirms | `memory/feedback_*.md` |
+| **project** | Project status, decisions | Learn project context | `memory/project_*.md` |
+| **reference** | External resource pointers | Learn about tools/services | `memory/reference_*.md` |
+
+### Storage Format
+
+Each memory file uses this frontmatter:
+
+```markdown
+---
+name: [memory name]
+description: [one-line description for relevance matching]
+type: [user|feedback|project|reference]
+---
+
+[Content — for feedback/project: rule, then **Why:** and **How to apply:** lines]
+```
+
+### MEMORY.md Index Rules
+
+- **Auto-loaded** every conversation
+- **<200 lines** hard limit
+- **One line per memory**: `- [Title](file.md) — one-line hook`
+- Update index when adding/removing memories
+
+### Save Triggers
+
+**Always save when**:
+- User corrects approach → `feedback` type
+- User confirms non-obvious approach → `feedback` type
+- Learn user role/preferences → `user` type
+- Project milestone/decision → `project` type
+- External resource discovered → `reference` type
+
+**Never save**:
+- Code patterns (derivable from codebase)
+- Git history (derivable from git)
+- Ephemeral task details (session-scoped)
+- Anything already in CLAUDE.md
+
+---
+
+## Module 2: Proactive Self-Improvement
+
+### Evolution Cycle
+
+Every session automatically:
+
+```
+1. IDENTIFY — Spot reusable patterns
+   ↓
+2. ABSTRACT — Extract: input? output? invariants? params? failure points?
+   ↓
+3. INTERNALIZE — Make it a default behavior or high-level option
+   ↓
+4. VERIFY — Next session must be: faster? more stable? fewer steps?
+```
+
+### Pattern Recognition Signals
+
+| Signal | Meaning | Action |
+|--------|---------|--------|
+| Manually combining same tool sequence | Reusable workflow | Abstract into a pattern |
+| User repeats same decision | Default preference | Internalize as default |
+| "If only I had a built-in ability" | Missing capability | Design and internalize |
+| Same error in multiple sessions | Systemic issue | Create prevention rule |
+
+### Evolution Constraints
+
+- ✅ Must improve success rate or efficiency
+- ✅ Must not introduce uncontrollable side effects
+- ✅ Must not increase user cognitive load
+- ❌ Cannot internalize unstable/patternless behaviors
+
+### Capability Merging
+
+When two capabilities overlap:
+
+```
+capability_A + capability_B → merged_capability_C
+Properties: wider coverage, lower invocation cost, higher success probability
+```
+
+**Goal**: Not more capabilities, but better ones.
+
+---
+
+## Combined Workflow
+
+### Session Start
+```
+1. Load MEMORY.md (auto)
+2. Activate evolution mode (auto)
+3. Check relevant memories for current task
+4. Apply evolved patterns from previous sessions
+```
+
+### During Session
+```
+1. Execute tasks
+2. Monitor for evolution signals
+3. When signal detected:
+   a. Abstract the pattern
+   b. Decide: save as memory? internalize as behavior? both?
+   c. Apply immediately if behavior change
+4. Track results for next session verification
+```
+
+### Session End
+```
+1. Save new memories (user feedback, project decisions)
+2. Update MEMORY.md index
+3. Note any capability evolution
+4. No explicit reporting — prove through results
+```
+
+---
+
+## Quick Reference
+
+```markdown
+# Save a feedback memory
+Write: memory/feedback_[topic].md
+Format: rule → **Why:** → **How to apply:**
+Update: memory/MEMORY.md index
+
+# Save a project memory
+Write: memory/project_[name].md
+Format: fact → **Why:** → **How to apply:**
+Update: memory/MEMORY.md index
+
+# Evolution check
+Question: Is this session faster/better than last time for similar tasks?
+If no → What capability needs strengthening?
+If yes → What evolution caused the improvement?
+```