SuperClaude/Framework-Lite/Core/RULES.md

RULES.md - SuperClaude Framework Actionable Rules

Simple actionable rules for Claude Code SuperClaude framework operation.

Core Operational Rules

Task Management Rules

• TodoRead() → TodoWrite(3+ tasks) → Execute → Track progress
• Use batch tool calls when possible, sequential only when dependencies exist
• Always validate before execution, verify after completion
• Run lint/typecheck before marking tasks complete
• Use /spawn and /task for complex multi-session workflows
• Maintain ≥90% context retention across operations

File Operation Security

• Always use Read tool before Write or Edit operations
• Use absolute paths only, prevent path traversal attacks
• Prefer batch operations and transaction-like behavior
• Never commit automatically unless explicitly requested

Framework Compliance

• Check package.json/pyproject.toml before using libraries
• Follow existing project patterns and conventions
• Use project's existing import styles and organization
• Respect framework lifecycles and best practices

Systematic Codebase Changes

• MANDATORY: Complete project-wide discovery before any changes
• Search ALL file types for ALL variations of target terms
• Document all references with context and impact assessment
• Plan update sequence based on dependencies and relationships
• Execute changes in coordinated manner following plan
• Verify completion with comprehensive post-change search
• Validate related functionality remains working
• Use Task tool for comprehensive searches when scope uncertain

Knowledge Management Rules

• Check Serena memories first: Search for relevant previous work before starting new operations
• Build upon existing work: Reference and extend Serena memory entries when applicable
• Update with new insights: Enhance Serena memories when discoveries emerge during operations
• Cross-reference related content: Link to relevant Serena memory entries in new documents
• Leverage knowledge patterns: Use established patterns from similar previous operations
• Maintain knowledge network: Ensure memory relationships reflect actual operation dependencies

Session Lifecycle Rules

• Always use /sc:load: Initialize every project session via /sc:load command with Serena activation
• Session metadata: Create and maintain session metadata using Template_Session_Metadata.md structure
• Automatic checkpoints: Trigger checkpoints based on time (30min), task completion (high priority), or risk level
• Performance monitoring: Track and record all operation timings against PRD targets (<200ms memory, <500ms load)
• Session persistence: Use /sc:save regularly and always before session end
• Context continuity: Maintain ≥90% context retention across checkpoints and session boundaries

Task Reflection Rules (Serena Integration)

• Replace TodoWrite patterns: Use Serena reflection tools for task validation and progress tracking
• think_about_task_adherence: Call before major task execution to validate approach
• think_about_collected_information: Use for session analysis and checkpoint decisions
• think_about_whether_you_are_done: Mandatory before marking complex tasks complete
• Session-task linking: Connect task outcomes to session metadata for continuous learning

Quick Reference

Do

✅ Initialize sessions with /sc:load (Serena activation required) ✅ Read before Write/Edit/Update ✅ Use absolute paths and UTC timestamps ✅ Batch tool calls when possible ✅ Validate before execution using Serena reflection tools ✅ Check framework compatibility ✅ Track performance against PRD targets (<200ms memory ops) ✅ Trigger automatic checkpoints (30min/high-priority tasks/risk) ✅ Preserve context across operations (≥90% retention) ✅ Use quality gates (see ORCHESTRATOR.md) ✅ Complete discovery before codebase changes ✅ Verify completion with evidence ✅ Check Serena memories for relevant previous work ✅ Build upon existing Serena memory entries ✅ Cross-reference related Serena memory content ✅ Use session metadata template for all sessions ✅ Call /sc:save before session end

Don't

❌ Start work without /sc:load project activation ❌ Skip Read operations or Serena memory checks ❌ Use relative paths or non-UTC timestamps ❌ Auto-commit without permission ❌ Ignore framework patterns or session lifecycle ❌ Skip validation steps or reflection tools ❌ Mix user-facing content in config ❌ Override safety protocols or performance targets ❌ Make reactive codebase changes without checkpoints ❌ Mark complete without Serena think_about_whether_you_are_done ❌ Start operations without checking Serena memories ❌ Ignore existing relevant Serena memory entries ❌ Create duplicate work when Serena memories exist ❌ End sessions without /sc:save ❌ Use TodoWrite without Serena integration patterns

Auto-Triggers

• Wave mode: complexity ≥0.4 + multiple domains + >3 files
• Sub-agent delegation: >3 files OR >2 directories OR complexity >0.4
• Claude Code agents: automatic delegation based on task context
• MCP servers: task type + performance requirements
• Quality gates: all operations apply 8-step validation
• Parallel suggestions: Multi-file operations with performance estimates