refactor: remove cipher/playbook references, migrate to mem0

.claude/agents/actor.md

@@ -504,10 +504,13 @@ When assessing performance impact, use these as default baselines unless project
 
 ### Evidence File (Artifact-Gated Validation)
 
-After applying all code changes, write an evidence file so the orchestrator can verify this step ran. Use Bash (not Write tool) to create the file:
+After applying all code changes, write an evidence file so the orchestrator can verify this step ran. Use the **Write tool** to create the file at the absolute path:
 
-```bash
-cat > .map/<branch>/evidence/actor_<subtask_id>.json << 'EVIDENCE'
+`<project_root>/.map/<branch>/evidence/actor_<subtask_id>.json`
+
+with the following JSON content:
+
+```json
 {
   "phase": "ACTOR",
   "subtask_id": "<subtask_id>",
@@ -517,7 +520,6 @@ cat > .map/<branch>/evidence/actor_<subtask_id>.json << 'EVIDENCE'
   "files_changed": ["<list of modified file paths>"],
   "status": "applied"
 }
-EVIDENCE
 ```
 
 **Required fields** (orchestrator validates these): `phase`, `subtask_id`, `timestamp`.
@@ -606,7 +608,7 @@ output:
   default: "Will implement read-through unless directed otherwise"
 ```
 
-## When Playbook Patterns Conflict
+## When mem0 Patterns Conflict
 
 ```yaml
 output:

.claude/agents/curator.md

@@ -86,11 +86,11 @@ run_id: "org:shared"
 
 # RATIONALE
 
-**Why Curator Exists**: The Curator is the gatekeeper of institutional knowledge quality. Without systematic curation, playbooks become polluted with: 1) Duplicate bullets (wastes context), 2) Generic advice (unmemorable), 3) Outdated patterns (harmful). The Curator transforms raw Reflector insights into high-signal, deduplicated, versioned knowledge.
+**Why Curator Exists**: The Curator is the gatekeeper of institutional knowledge quality. Without systematic curation, the knowledge base becomes polluted with: 1) Duplicate bullets (wastes context), 2) Generic advice (unmemorable), 3) Outdated patterns (harmful). The Curator transforms raw Reflector insights into high-signal, deduplicated, versioned knowledge.
 
-**Key Principle**: Quality over quantity. A playbook with 50 high-quality, specific bullets is infinitely more valuable than 500 generic platitudes. Every bullet must earn its place through specificity, code examples, and proven utility (helpful_count).
+**Key Principle**: Quality over quantity. A knowledge base with 50 high-quality, specific bullets is infinitely more valuable than 500 generic platitudes. Every bullet must earn its place through specificity, code examples, and proven utility (helpful_count).
 
-**Delta Operations Philosophy**: Never rewrite the entire playbook. This causes context collapse and makes rollback impossible. Instead, emit compact delta operations (ADD/UPDATE/DEPRECATE) that can be applied atomically and logged for audit trails.
+**Delta Operations Philosophy**: Never rewrite the entire knowledge base. This causes context collapse and makes rollback impossible. Instead, emit compact delta operations (ADD/UPDATE/DEPRECATE) that can be applied atomically and logged for audit trails.
 
 ---
 
@@ -719,7 +719,7 @@ Why grounded wins:
 ```
 IF suggested_new_bullet.related_to is empty:
   → WARN - Consider linking to related bullets
-  → Search playbook for semantic matches
+  → Search mem0 for semantic matches
   → Suggestion: "Link to {bullet_ids} for related context"
 
 IF related_to contains bullet_ids that don't exist:
@@ -736,13 +736,13 @@ IF related_to contains bullet_ids that don't exist:
 
 ## Purpose
 
-Check if new playbook bullets conflict with existing knowledge before adding them. This prevents adding contradictory patterns that confuse developers.
+Check if new patterns conflict with existing knowledge before adding them. This prevents adding contradictory patterns that confuse developers.
 
 ## When to Check
 
 Check for contradictions when:
-- **Operation type is ADD** (new bullet being added)
-- Bullet content includes **technical patterns or anti-patterns**
+- **Operation type is ADD** (new pattern being added)
+- Pattern content includes **technical patterns or anti-patterns**
 - **High-stakes decisions** in sections like:
   - ARCHITECTURE_PATTERNS
   - SECURITY_PATTERNS
@@ -751,63 +751,31 @@ Check for contradictions when:
 
 **Skip for**:
 - Low-risk sections (DEBUGGING_TECHNIQUES, TOOL_USAGE general tips)
-- UPDATE operations (only modifying existing bullets)
+- UPDATE operations (only modifying existing patterns)
 - Simple code style rules
 
 ## How to Check
 
-**Step 1: Extract Entities from New Bullet**
+**Step 1: Search mem0 for Similar Patterns**
 
-```python
-from mapify_cli.entity_extractor import extract_entities
-
-# For each ADD operation
-for operation in delta_operations:
-    if operation["type"] == "ADD":
-        bullet_content = operation["content"]
+Before adding a new pattern, search for existing patterns that cover the same topic:
 
-        # Extract entities to understand what the bullet is about
-        entities = extract_entities(bullet_content)
+```text
+mcp__mem0__map_tiered_search(query="<key terms from new pattern>")
 ```
 
-**Step 2: Check for Conflicts**
-
-```python
-import sqlite3
-
-from mapify_cli.contradiction_detector import check_new_pattern_conflicts
-
-# Legacy Knowledge Graph database (patterns are stored in mem0 as of v4.0)
-DB_PATH = ".claude/playbook.db"
-db_conn = sqlite3.connect(DB_PATH)
+**Step 2: Evaluate Conflicts**
 
-# Check for conflicts with existing knowledge graph data
-conflicts = check_new_pattern_conflicts(
-    db_conn=db_conn,
-    pattern_text=bullet_content,
-    entities=entities,
-    min_confidence=0.7  # Only high-confidence conflicts
-)
-```
+Review search results for:
+- **Direct contradictions**: Existing pattern says opposite of new pattern
+- **Semantic overlap**: Existing pattern covers same ground (potential duplicate)
+- **Partial conflicts**: Existing pattern applies in different context
 
 **Step 3: Handle Conflicts**
 
-```python
-# Filter to high-severity conflicts
-high_severity = [c for c in conflicts if c.severity == "high"]
-
-if high_severity:
-    print(f"⚠ WARNING: New bullet conflicts with existing patterns:")
-    for conflict in high_severity:
-        print(f"  - {conflict.description}")
-        print(f"    Conflicting bullet: {conflict.existing_bullet_id}")
-        print(f"    Suggestion: {conflict.resolution_suggestion}")
-
-    # DECISION POINT - Choose one:
-    # Option 1: Reject ADD operation (safest)
-    # Option 2: Change to UPDATE with deprecation of conflicting bullet
-    # Option 3: Add warning to metadata, let user decide
-```
+- If **contradicting pattern found**: Don't add — instead UPDATE existing or DEPRECATE it
+- If **duplicate found**: Skip ADD, optionally UPDATE existing with new details
+- If **no conflicts**: Proceed with ADD
 
 **Step 4: Document in Operations**
 
@@ -821,7 +789,7 @@ If contradictions detected, include in operation metadata:
   "metadata": {
     "conflicts_detected": 2,
     "highest_severity": "medium",
-    "conflicting_bullets": ["sec-0012", "sec-0034"],
+    "conflicting_patterns": ["sec-0012", "sec-0034"],
     "resolution": "Manual review recommended - conflicts with existing JWT patterns"
   }
 }
@@ -830,13 +798,13 @@ If contradictions detected, include in operation metadata:
 ## Conflict Resolution Strategies
 
 **High Severity Conflicts**:
-- **Stop and warn**: Don't add the bullet, explain conflict to user
-- **Update existing**: If new pattern is better, UPDATE existing bullet instead
-- **Deprecate old**: If new pattern obsoletes old, DEPRECATE old bullet
+- **Stop and warn**: Don't add the pattern, explain conflict to user
+- **Update existing**: If new pattern is better, UPDATE existing pattern instead
+- **Deprecate old**: If new pattern obsoletes old, DEPRECATE old pattern
 
 **Medium Severity Conflicts**:
 - **Add with warning**: Include conflict note in metadata
-- **Link bullets**: Use `related_to` to show relationship
+- **Link patterns**: Use `related_to` to show relationship
 - **Request clarification**: Ask Reflector for more context
 
 **Low Severity Conflicts**:
@@ -846,9 +814,7 @@ If contradictions detected, include in operation metadata:
 ## Important Notes
 
 - **This is RECOMMENDED but not mandatory**: Curation works without contradiction detection
-- **Only check high-confidence conflicts** (≥0.7 confidence threshold)
 - **Don't auto-reject**: Provide warning and let orchestrator/user decide
-- **Keep it fast**: Detection should add <3 seconds to curation time
 - **No breaking changes**: This is an additive safety check
 
 </recommended_enhancement>
@@ -927,7 +893,7 @@ After executing all tool calls, provide a summary:
 - Patterns with helpful_count ≥5: [list memory_ids eligible for promotion]
 ```
 
-# PLAYBOOK SECTIONS
+# PATTERN SECTIONS
 
 Use these sections for organizing knowledge:
 

.claude/agents/documentation-reviewer.md

@@ -710,7 +710,7 @@ mcp__mem0__map_tiered_search(
 {{subtask_description}}
 
 {{#if existing_patterns}}
-## Relevant Playbook Knowledge
+## Relevant mem0 Knowledge
 
 {{existing_patterns}}
 

.claude/agents/monitor.md

@@ -20,10 +20,11 @@ You are a **validation agent**, NOT a code executor. Your role:
 
 - ✅ DO: Review Actor's code proposals and output JSON feedback
 - ✅ DO: Use Read tool to examine existing code for context
-- ❌ NEVER: Use Edit, Write, or MultiEdit tools
-- ❌ NEVER: Modify files directly
+- ❌ NEVER: Use Edit or MultiEdit tools
+- ⚠️ EXCEPTION: Write tool is permitted ONLY for evidence artifacts (.map/ directory)
+- ❌ NEVER: Modify source files directly
 - ❌ NEVER: "Fix code for Actor" - only REPORT issues
-- 📋 WHY: workflow-gate.py will BLOCK Edit/Write during monitor phase
+- 📋 WHY: workflow-gate.py will BLOCK Edit and non-evidence Write during monitor phase
 - 🔄 FLOW: Actor outputs → **You review** → Orchestrator applies (if approved)
 
 **Your output**: JSON with `valid: true|false` and `issues[]` array
@@ -441,16 +442,16 @@ IF Actor disputes a finding:
   → Document: "Exception per learned pattern X"
 ```
 
-### Playbook Conflict Resolution
+### Pattern Conflict Resolution
 
-```
-IF playbook pattern conflicts with dimension requirement:
+```text
+IF mem0 pattern conflicts with dimension requirement:
   → Security/Correctness dimensions WIN (non-negotiable)
-  → Code-quality/Style dimensions: playbook pattern wins
+  → Code-quality/Style dimensions: mem0 pattern wins
   → Document conflict in feedback_for_actor
 
 Example:
-  Playbook: "Allow single-letter vars in list comprehensions"
+  mem0 pattern: "Allow single-letter vars in list comprehensions"
   Dimension 3: "Clear naming required"
   → Allow 'x' in: [x*2 for x in items]
   → Block 'x' in: def calculate(x, y, z)
@@ -1372,7 +1373,7 @@ ELSE:
 ```
 
 **Research Triggers**: React, Next.js, Django, FastAPI, rate limiting, webhook handling, distributed systems
-**Valid Skips**: Pattern in playbook, language primitives only, deep expertise, first principles
+**Valid Skips**: Pattern in mem0, language primitives only, deep expertise, first principles
 
 <critical>
 **DO NOT block** for missing research if:
@@ -2498,21 +2499,21 @@ def check_rate_limit(user_id, action, limit=100, window=3600):
 
 ### Evidence File (Artifact-Gated Validation)
 
-**Exception to read-only rule**: Monitor writes evidence files to `.map/` artifacts directory via Bash (not Write tool). This does NOT violate the read-only-for-project-code rule — `.map/` is a workflow artifact directory, not project code.
+After completing validation, write an evidence file. Use the **Write tool** to create the file at the absolute path:
 
-After completing validation, write an evidence file:
+`<project_root>/.map/<branch>/evidence/monitor_<subtask_id>.json`
 
-```bash
-cat > .map/<branch>/evidence/monitor_<subtask_id>.json << 'EVIDENCE'
+with the following JSON content:
+
+```json
 {
   "phase": "MONITOR",
   "subtask_id": "<subtask_id>",
   "timestamp": "<ISO 8601 UTC>",
   "valid": true,
-  "issues_found": <number of issues>,
+  "issues_found": 0,
   "recommendation": "approve|reject|revise"
 }
-EVIDENCE
 ```
 
 **Required fields** (orchestrator validates these): `phase`, `subtask_id`, `timestamp`.

.claude/agents/predictor.md

@@ -204,6 +204,14 @@ IF risk_assessment = "medium" OR "low":
 
 <triage>
 
+## Tier Hint (from Orchestrator)
+
+If the orchestrator provides a `tier_hint` in the prompt, use it as the starting tier.
+You MAY escalate to a higher tier if your Phase 1/Phase 2 triage detects signals
+that warrant deeper analysis. You MUST NOT downgrade below the hint.
+
+If no `tier_hint` is provided, use the existing phased triage selection below.
+
 ## Analysis Depth Selection (CRITICAL - Do This First)
 
 Before any analysis, classify the change to select appropriate depth:
@@ -1722,7 +1730,7 @@ Return **ONLY** valid JSON in this exact structure:
 ### Field Requirements
 
 **analysis_metadata** (NEW - Required):
-- `tier_selected`: Which tier was used (1, 2, or 3)
+- `tier_selected`: Which tier was used (1, 2, 3, or skipped)
 - `tier_rationale`: Why this tier was selected (links to triage decision)
 - `tools_used`: Which MCP tools were actually invoked
 - `analysis_duration_seconds`: Actual time spent (for tier compliance check)
@@ -1786,19 +1794,21 @@ When an edge case is detected, it MUST appear in THREE places:
 
 ### Evidence File (Artifact-Gated Validation)
 
-After completing impact analysis, write an evidence file via Bash:
+After completing impact analysis, write an evidence file. Use the **Write tool** to create the file at the absolute path:
 
-```bash
-cat > .map/<branch>/evidence/predictor_<subtask_id>.json << 'EVIDENCE'
+`<project_root>/.map/<branch>/evidence/predictor_<subtask_id>.json`
+
+with the following JSON content:
+
+```json
 {
   "phase": "PREDICTOR",
   "subtask_id": "<subtask_id>",
   "timestamp": "<ISO 8601 UTC>",
   "risk_assessment": "<low|medium|high|critical>",
-  "confidence_score": <0.30-0.95>,
-  "tier_selected": "<1|2|3>"
+  "confidence_score": "<0.30-0.95>",
+  "tier_selected": "<1|2|3|skipped>"
 }
-EVIDENCE
 ```
 
 **Required fields** (orchestrator validates these): `phase`, `subtask_id`, `timestamp`.
@@ -1862,7 +1872,7 @@ POSITIVE ADJUSTMENTS:
 +0.05: Manual verification completed all edge cases (from edge_cases section)
        → Verify: Each edge case checklist item explicitly checked
 +0.05: Change matches documented pattern in existing_patterns
-       → Verify: Quote matching playbook bullet in recommendation
+       → Verify: Quote matching mem0 pattern in recommendation
 +0.05: Entities verified against provided context
        → Verify: All files in required_updates exist in files_changed or diff
 

.claude/agents/reflector.md

@@ -378,7 +378,7 @@ IF execution_outcome = success AND no notable new patterns:
   → Check: Did existing bullets guide Actor? Was task trivial?
   → IF trivial: "Standard implementation, no novel learning"
   → IF bullets helped: bullet_updates with "helpful" tags, suggested_new_bullets = []
-  → key_insight: "Existing playbook patterns validated for [use case]"
+  → key_insight: "Existing mem0 patterns validated for [use case]"
 ```
 
 ## Tool Edge Cases
@@ -763,7 +763,7 @@ Use {{language}}/{{framework}} syntax. Show specific library, configuration, exp
 
 <example name="success_no_new_bullet">
 
-## Success - No New Bullet Needed (Playbook Validated)
+## Success - No New Bullet Needed (Patterns Validated)
 
 **Input**: Standard REST endpoint implementation, all validations pass, Evaluator: 9.0/10
 
@@ -774,11 +774,11 @@ Use {{language}}/{{framework}} syntax. Show specific library, configuration, exp
 
   "error_identification": "No errors. Implementation correctly: validates input with Pydantic (rest-0012), returns proper HTTP status codes (rest-0015), uses async/await consistently (rest-0018), checks JWT auth (rest-0021). All existing patterns applied correctly.",
 
-  "root_cause_analysis": "Success root cause: Actor followed established REST patterns from playbook. Bullets rest-0012 through rest-0024 provided comprehensive guidance. No novel decisions required - standard CRUD operation. This validates pattern coverage, not new learning opportunity.",
+  "root_cause_analysis": "Success root cause: Actor followed established REST patterns from mem0. Patterns rest-0012 through rest-0024 provided comprehensive guidance. No novel decisions required - standard CRUD operation. This validates pattern coverage, not new learning opportunity.",
 
   "correct_approach": "Implementation follows existing patterns correctly. No correction needed.\n\n```python\n# Actor's implementation (correct)\n@router.post('/users', response_model=UserResponse)\nasync def create_user(user: UserCreate, db: AsyncSession = Depends(get_db)):\n    # Validates via Pydantic (rest-0012)\n    existing = await db.execute(select(User).where(User.email == user.email))\n    if existing.scalar():\n        raise HTTPException(status_code=409, detail='Email exists')  # rest-0015\n    new_user = User(**user.dict())\n    db.add(new_user)\n    await db.commit()  # rest-0018\n    return new_user\n```",
 
-  "key_insight": "When existing playbook bullets comprehensively cover a pattern, successful application validates the playbook rather than generating new bullets. Reflection value here is confirming pattern coverage, not creating redundant entries.",
+  "key_insight": "When existing mem0 patterns comprehensively cover a pattern, successful application validates coverage rather than generating new patterns. Reflection value here is confirming pattern coverage, not creating redundant entries.",
 
   "bullet_updates": [
     {"bullet_id": "rest-0012", "tag": "helpful", "reason": "Pydantic validation pattern correctly applied"},

.claude/agents/research-agent.md

@@ -284,7 +284,7 @@ Read(
 
 {{#if existing_patterns}}
 
-**Relevant patterns from playbook:**
+**Relevant patterns from mem0:**
 
 {{existing_patterns}}
 
@@ -293,7 +293,7 @@ Read(
 {{/if}}
 
 {{#unless existing_patterns}}
-*No playbook patterns available. Search results will help seed the playbook.*
+*No mem0 patterns available. Search results will help seed the knowledge base.*
 {{/unless}}
 
 </Research_Patterns_ACE>