Wolf's 10 core principles for agent behavior and system design
This skill provides access to Wolf's 10 core principles that guide the design, implementation, and operation of the Wolf Agents multi-agent system. These principles have been refined over 50+ phases of real-world development.
Principle: All work produces durable, verifiable artifacts rather than ephemeral conversations.
Implementation:
Example Application:
Instead of: "I fixed the bug, it works now"
Do this: Create PR with fix, tests, and documentation of root cause
Principle: Each agent role has clearly defined responsibilities with minimal overlap and strict boundaries.
Implementation:
Example Application:
PM Agent: Defines requirements and acceptance criteria
Coder Agent: Implements solution meeting criteria
Reviewer Agent: Validates implementation quality
QA Agent: Verifies functionality and tests
Principle: All implementation work must be preceded by structured research and evidence-based recommendations. This applies at TWO levels:
Implementation:
Level 1 - Architectural Research:
research label as a blocking gate for implementationLevel 2 - Documentation Lookup:
Why Two Levels:
Example Application:
Task: Add authentication to API
Level 1 - Architectural Research (research-agent, 4 hours):
- Analyze existing auth patterns, security requirements, compliance needs
- Compare JWT vs OAuth2 vs Passport.js approaches
- Evaluate security implications, scalability, maintenance burden
- Deliver recommendation: "Use Passport.js with JWT strategy"
→ Output: ADR documenting decision and rationale
Level 2 - Documentation Lookup (coder-agent, 3 minutes):
- WebSearch "passport.js jwt strategy official documentation 2025"
- WebFetch https://www.passportjs.org/packages/passport-jwt/
- Verify: Current version is 4.0.1, check for breaking changes from 3.x
- Review: Example code for JWT verification and token extraction
→ Output: Implementation using current, verified API patterns
Result: Implementation informed by both architectural research (Level 1)
and current documentation (Level 2), avoiding both strategic and tactical errors.
Principle: New policies and constraints are tested in advisory mode before becoming hard gates.
Implementation:
Example Application:
New Rule: All PRs must have 90% test coverage
Phase 1: Report coverage but don't block (2 weeks)
Phase 2: Block if <70% coverage (2 weeks)
Phase 3: Enforce 90% threshold (ongoing)
Principle: All decisions must be supported by concrete evidence and measurable outcomes.
Implementation:
Example Application:
Decision: Choose between REST and GraphQL
Evidence Required:
- Latency benchmarks for typical queries
- Bundle size impact measurements
- Developer productivity metrics
- Maintenance cost analysis
Principle: The system continuously learns from its operations and evolves based on evidence.
Implementation:
Example Application:
Problem: CI failures increasing
Journal: Document failure patterns
Analysis: Identify common root causes
Improvement: Add pre-commit checks for identified patterns
Measurement: Track CI failure rate reduction
Principle: The system must operate reliably across multiple AI providers with graceful fallback.
Implementation:
Example Application:
Primary: OpenAI GPT-4 for complex reasoning
Fallback 1: Claude for continued operation
Fallback 2: Local models for basic functionality
Circuit Breaker: Automatic switching based on availability
Principle: Leverage GitHub platform primitives to minimize custom infrastructure and operational overhead.
Implementation:
Example Application:
Instead of: Custom task tracking system
Use: GitHub Issues with labels and milestones
Instead of: Custom CI/CD pipeline
Use: GitHub Actions with reusable workflows
Principle: All work should be broken into small, independently valuable increments.
Implementation:
Example Application:
Feature: User Dashboard
Increment 1: Basic layout and navigation (2h)
Increment 2: User profile widget (3h)
Increment 3: Activity feed (4h)
Increment 4: Settings panel (2h)
Each increment is fully functional and deployable
Principle: All decisions, processes, and constraints must be openly documented and auditable.
Implementation:
Example Application:
Decision: Change deployment frequency
Documentation: ADR with rationale
Audit Trail: Git history of decision
Review: Monthly governance meeting
Update: Adjust based on operational metrics
You can ask about specific principles or search across all principles:
"What is principle 5?" → Returns Evidence-Based Decision Making
"How does Wolf handle security?" → Returns relevant principles (2, 5, 7)
"How to make decisions?" → Returns principles 3, 5, 6
"Show all principles" → Returns complete list with summaries
When principles appear to conflict, use this priority order:
query.js - Search principles by ID, keyword, or topicapply.js - Generate principle-based recommendations for specific scenariosThese principles evolve based on operational evidence. Changes require:
If you catch yourself thinking:
STOP. Use Skill tool to load wolf-principles BEFORE proceeding.
REQUIRED NEXT STEPS:
Sequential skill chain - DO NOT skip steps
REQUIRED NEXT SKILL: Use wolf-archetypes to determine behavioral archetype
REQUIRED NEXT SKILL: Use wolf-governance to identify quality gates
REQUIRED NEXT SKILL: Use wolf-roles to understand collaboration patterns
DO NOT PROCEED to implementation without completing steps 1-3.
Before claiming you've applied principles:
Can't check all boxes? Work is incomplete. Return to this skill.
Source: docs/principles.md (lines 292-527) Last Updated: 2025-11-14 Phase: Superpowers Skill-Chaining Enhancement v2.0.0
