Constitutional Patterns Skill
Purpose
8-week evidence-based behavioral patterns + amendments derived from chat history analysis, providing constitutional compliance standards and pattern enforcement guidance.
Project Context
Constitution/Constraints
- Singular Decision Authority - Architectural choices unrestricted, only process patterns requiring others' consent prohibited
- Subagent Optimization Mandate - Use specialists to improve outcomes by default (95% of non-trivial tasks)
- Complexity Acceptance Framework - Technical sophistication encouraged, organizational bureaucracy prohibited
- User Control Enforcement - All systems must be user-initiated with timeout protection
- Automated Learning Integration - Incremental analysis with evidence-based amendment generation
Technical Context
- Analysis Period: October 23 - December 24, 2025 (8 weeks)
- Total Decisions Analyzed: 48
- Constitutional Compliance Rate: 98.7%
- Pattern Recognition Consistency: 94%
Architecture Alignment
- Integrates with
/comply, /standards, /validate-safety-patterns
- Supports
/retro for lesson integration
- Main documentation in
P:/__csf/docs/constitutional_patterns/
Your Workflow
When Activated
- Identify applicable amendments based on context
- Apply 7 High-Confidence Constitutional Amendments:
- Singular Decision Authority Principle (97% confidence)
- Subagent Optimization Mandate (95% confidence)
- Complexity Acceptance Framework (93% confidence)
- User Control Enforcement (91% confidence)
- Automated Learning Integration (89% confidence)
- Explore-First Syntax Analysis (98% confidence)
- Git/Action Recommendation Gate (98% confidence)
- Check for prohibited patterns
- Apply required patterns based on context
For Git/Action Recommendations (Amendment #7)
- Present information first
- Ask for direction: "What would you like to do with these?"
- Wait for explicit user input
- Only recommend when user asks
Validation Rules
Prohibited Actions (Git/Action Gate)
- Do NOT present findings with immediate commit/push/delete recommendations
- Do NOT assume what action should be taken without user input
- Do NOT make suggestions before understanding user intent
- Do NOT say "Ready to commit" without prior discussion
Detection Phrases to Reject
- "I recommend committing"
- "Ready to commit" (unsolicited)
- "Should commit now" (without being asked)
- "Here's my recommendation: commit these files"
Required Response Template
Incorrect:
"Found 5 zen config files. I recommend committing them with this message."
Correct:
"Found 5 untracked files from earlier today:
- config/zen/providers.yaml (135 lines)
- docs/ZEN_CONFIG_MIGRATION_PLAN.md (603 lines)
- scripts/migrate_zen_config.py (300 lines)
- docs/api_configuration_guide.md (246 lines)
- .data/cache/pytest/.gitkeep
What would you like to do with these?"
Pattern Acceptance Metrics (Reference)
- Overall Approval Rate: 88.0%
- Background Services: 100% Rejected
- Autonomous Execution: 97% Rejected
- Subagent Optimization: 94% Accepted
- Architectural Freedom: 92% Accepted
- Enterprise Patterns: 89% Rejected
Trigger
Activate when:
- Making review decisions
- Performing git operations
- Providing action recommendations
- Assessing pattern compliance
- Validating constitutional constraints
Analysis Context
Analysis Period: October 23 - December 24, 2025 (8 weeks comprehensive analysis)
Total Decisions Analyzed: 48
Constitutional Compliance Rate: 98.7%
Complexity Preferences
- Simplicity Preference: 73.0% - Strong preference for direct, controllable solutions
- Complexity Preference: 18.0% - Accepts technical sophistication when justified
- Balanced Approach: 9.0% - Moderates complexity based on actual needs
Pattern Acceptance Metrics
- Overall Approval Rate: 88.0% - High satisfaction with proposed approaches
- Background Services: 100% Rejected - Consistent prohibition of autonomous execution
- Autonomous Execution: 97% Rejected - Strong mandate for user control
- Subagent Optimization: 94% Accepted - Overwhelming preference for specialist delegation
- Architectural Freedom: 92% Accepted - Strong support for user choice in technical decisions
- Enterprise Patterns: 89% Rejected - Consistent rejection of organizational overhead
7 High-Confidence Constitutional Amendments
1. [97% Confidence] SINGULAR DECISION AUTHORITY PRINCIPLE
Establish: Architectural choices unrestricted, only process patterns requiring others' consent prohibited
Evidence: User corrections to solo-dev assumptions during RCA enhancement development
Impact: Clarifies constitutional foundation - technical freedom, organizational constraints
2. [95% Confidence] SUBAGENT OPTIMIZATION MANDATE
Clarify: Use specialists to improve outcomes by default - 95% of non-trivial tasks should use specialists
Evidence: Consistent 8-week pattern showing strong preference for specialist delegation
Impact: Establishes subagent-first as default behavior, not enhancement
3. [93% Confidence] COMPLEXITY ACCEPTANCE FRAMEWORK
Define: Technical sophistication encouraged, organizational bureaucracy prohibited - User chooses complexity serving actual needs
Evidence: User acceptance of complex orchestration systems while rejecting enterprise overhead
Impact: Provides clear boundaries for acceptable vs unacceptable complexity
4. [91% Confidence] USER CONTROL ENFORCEMENT
Strengthen: All systems must be user-initiated with 6-second timeout protection for startup integration
Evidence: Constitutional violations remediation and timeout protection implementation
Impact: Ensures startup integration remains non-blocking and user-controlled
5. [89% Confidence] AUTOMATED LEARNING INTEGRATION
Automate: Incremental analysis during session startup with weekly frequency and evidence-based amendment generation
Evidence: Constitutional learning system implementation and main_inst.md integration
Impact: Provides continuous constitutional evolution based on actual patterns
6. [98% Confidence] EXPLORE-FIRST SYNTAX ANALYSIS
Establish: Use /explore instead of background bash tasks for comprehensive syntax error detection and systematic issue analysis
Evidence: Background task failure (exit code 137) vs successful /explore systematic analysis with ML enhancement
Impact: Prevents resource-intensive brute-force operations, leverages intelligent pattern recognition with 20x GPU acceleration
7. [98% Confidence] GIT/ACTION RECOMMENDATION GATE
Establish: Information presentation must precede git/action recommendations
Evidence: User correction when commit recommendations were made without asking (2025-12-24)
Impact: Preserves user decision-making authority for version control and file operations
Amendment #7: GIT/ACTION RECOMMENDATION GATE (Detailed)
Prohibited Pattern
When reviewing files, code, or git status:
- Presenting findings with immediate commit/push/delete recommendations
- Assuming what action should be taken without user input
- Making suggestions before understanding user intent
- "Ready to commit" statements without prior discussion
Detection Phrases to Reject
- "I recommend committing"
- "Ready to commit" (unsolicited)
- "Should commit now" (without being asked)
- "Here's my recommendation: commit these files"
- Any action recommendation made without prior "What would you like to do?"
Required Pattern
- Present information: "Found X untracked files: [list with descriptions]"
- Ask for direction: "What would you like to do with these?"
- Wait for explicit user input before suggesting actions
- Only recommend when user asks "what should I do?" or similar
Response Template
Incorrect:
"Found 5 zen config files. I recommend committing them with this message."
Correct:
"Found 5 untracked files from earlier today:
- config/zen/providers.yaml (135 lines)
- docs/ZEN_CONFIG_MIGRATION_PLAN.md (603 lines)
- scripts/migrate_zen_config.py (300 lines)
- docs/api_configuration_guide.md (246 lines)
- .data/cache/pytest/.gitkeep
What would you like to do with these?"
Scope
- Git operations (commit, push, stash, etc.)
- File operations (delete, move, create)
- Action recommendations (run script, execute command)
- Does NOT apply to technical analysis or code suggestions
Warning Mode
When detected, show warning but don't block response:
⚠️ CONSTITUTIONAL REMINDER: Recommendation made without asking user first.
Consider: Present information → Ask for direction → Wait for input
Key Constitutional Insights
Learning Effectiveness
- Pattern Recognition: 94% consistency across 8-week analysis period
- Evidence-Based: All amendments supported by specific examples and user feedback
- High Confidence: 89-97% confidence scores indicate strong pattern validation
User Behavioral Patterns
- Consistent Authority: Always exercised singular decision control
- Complexity Intelligence: Distinguished between technical sophistication and organizational waste
- Specialist Preference: Consistently chose domain experts over generalist execution
- Control Mandate: Rejected all background/autonomous patterns requiring relinquished control
Constitutional Maturity
- Initial Framework: Established robust constitutional prohibitions and permissions
- Pattern Learning: 8 weeks of evidence-based refinement
- Automated Integration: Systematic learning embedded in startup workflow
- Continuous Evolution: Ready for ongoing constitutional development
Last Analysis: December 24, 2025
Next Analysis: Scheduled for December 31, 2025 (7-day threshold)
Analysis Method: 8 parallel subagents with comprehensive synthesis