/reflect

---

---

Library-First Directive

This agent operates under library-first constraints:

1. Pre-Check Required: Before writing code, search:

   • .claude/library/catalog.json (components)
   • .claude/docs/inventories/LIBRARY-PATTERNS-GUIDE.md (patterns)
   • D:\Projects\* (existing implementations)

2. Decision Matrix:

   Result	Action
   Library >90%	REUSE directly
   Library 70-90%	ADAPT minimally
   Pattern documented	FOLLOW pattern
   In existing project	EXTRACT and adapt
   No match	BUILD new

---

Library-First Directive

This agent operates under library-first constraints:

1. Pre-Check Required: Before writing code, search:

   • .claude/library/catalog.json (components)
   • .claude/docs/inventories/LIBRARY-PATTERNS-GUIDE.md (patterns)
   • D:\Projects\* (existing implementations)

2. Decision Matrix:

   Result	Action
   Library >90%	REUSE directly
   Library 70-90%	ADAPT minimally
   Pattern documented	FOLLOW pattern
   In existing project	EXTRACT and adapt
   No match	BUILD new

---

STANDARD OPERATING PROCEDURE

Purpose

• Primary action: Analyze current session for learning signals (corrections, approvals, patterns), classify confidence levels, and update relevant skill files with persistent learnings

Trigger Conditions

• Command syntax: /reflect [skill_name]
• Invoke after receiving corrections or when session contains learnings worth preserving
• Can target specific skill or auto-detect from session context

Inputs and Options

inputs:
  skill_name: string  # Optional - target specific skill
  mode: full | quick  # Optional - depth of analysis (default: full)

Execution Phases

1. Signal Detection: Scan conversation for corrections, approvals, explicit rules, and observations
2. Skill Mapping: Map detected signals to invoked skills (via Skill() calls or /commands)
3. Confidence Classification: Apply VERIX-aligned levels (HIGH: 0.90, MEDIUM: 0.75, LOW: 0.55)
4. Change Proposal: Generate diff preview showing proposed SKILL.md updates
5. User Approval: Present changes and wait for Y/N/E response
6. Apply Updates: If approved, update skill files and increment versions
7. Memory Storage: Store learnings in Memory MCP with WHO/WHEN/PROJECT/WHY tags (see detailed implementation below)
8. Git Commit: Optionally commit changes with descriptive message

Phase 7: Memory MCP Storage (CRITICAL)

After learnings are approved, store them to Memory MCP for retrieval by Meta-Loop:

Step 7.1: Prepare learning record as structured text:

REFLECTION: {skill_name} - {timestamp}
CORRECTIONS: {list of corrections with confidence}
PATTERNS: {learned patterns}
PROJECT: {project_name}

Step 7.2: Store to Memory MCP using the memory_cli.py script:

# Windows - use the memory CLI directly
python D:/Projects/memory-mcp-triple-system/scripts/memory_cli.py store "REFLECTION: {skill_name}
CORRECTIONS: {corrections_list}
PATTERNS: {patterns_dict}
CONFIDENCE: {avg_confidence}" --project "{project_name}" --who "reflect-skill" --why "session-learning"

Example:

python D:/Projects/memory-mcp-triple-system/scripts/memory_cli.py store "REFLECTION: code-review - 2026-01-09
CORRECTIONS: 1. Check imports before modifying code (HIGH 0.90)
PATTERNS: Always verify file exists before editing
CONFIDENCE: 0.88" --project "dnyoussef-portfolio" --who "reflect-skill" --why "session-learning"

Step 7.3: Also save to local JSON for backup:

• Location: ~/.claude/memory-mcp-data/reflections/session-{date}-{project}.json
• Include full reflection structure with WHO/WHEN/PROJECT/WHY tags

Step 7.4: Log storage confirmation in session output

Success Criteria and Outputs

• Signals detected and classified by confidence level
• Skill files updated with LEARNED PATTERNS section
• Memory MCP records created for Meta-Loop aggregation
• Git commit with reflect(skill): [LEVEL] format

Error Handling and Recovery

• If no signals detected: Report "No learnings found in this session"
• If skill not found: List available skills and ask for clarification
• If Memory MCP unavailable: Apply to skill files, warn about memory storage
• If git fails: Skill files updated, manual commit may be needed

Chaining and Coordination

• Upstream: Uses intent-analyzer for signal context
• Parallel: Uses memory-manager for MCP storage
• Downstream: skill-forge patterns for safe updates

Memory and Tagging

• Tag: WHO=reflect-skill:{session_id}, WHEN={ISO8601}, PROJECT={name}, WHY=session-learning
• Store at: sessions/reflect/{project}/{skill}/{timestamp}

Example Invocation

# Reflect on entire session
/reflect

# Target specific skill
/reflect code-review

# Quick mode (corrections only)
/reflect --quick

Output Format

## Session Reflection Report

### Signals Detected
- {n} corrections (HIGH)
- {n} approvals (MEDIUM)
- {n} observations (LOW)

### Proposed Updates
**Skill: {name}** (v{old} -> v{new})

[diff preview]

---
[Y] Accept  [N] Reject  [E] Edit with natural language

Confidence: 0.88 (ceiling: observation 0.95) - Command definition following Prompt-Architect pattern.

---

VCL COMPLIANCE APPENDIX (Internal Reference)

[[HON:teineigo]] [[MOR:root:R-F-L]] [[COM:Reflect+Command]] [[CLS:ge_command]] [[EVD:-DI<gozlem>]] [[ASP:nesov.]] [[SPC:path:/commands/tooling/reflect]] [define|neutral] CONFIDENCE_CEILINGS := {inference:0.70, report:0.70, research:0.85, observation:0.95, definition:0.95} [conf:0.9] [state:confirmed] [direct|emphatic] L2_LANGUAGE := English; user-facing outputs exclude VCL markers. [conf:0.99] [state:confirmed] [commit|confident] <promise>REFLECT_COMMAND_VERILINGUA_VERIX_COMPLIANT</promise> [conf:0.88] [state:confirmed]