Nice-Wolf-Studio

wolf-governance

2
0
# Install this skill:
npx skills add Nice-Wolf-Studio/wolf-skills-marketplace --skill "wolf-governance"

Install specific skill from multi-skill repository

# Description

Wolf's governance framework, compliance rules, quality gates, and process standards

# SKILL.md


name: wolf-governance
description: Wolf's governance framework, compliance rules, quality gates, and process standards
version: 1.2.0
triggers:
- "governance"
- "compliance"
- "quality gates"
- "process rules"
- "approval requirements"


Wolf Governance Skill

This skill provides access to Wolf's governance framework, including compliance requirements, quality gates, approval hierarchies, and process standards refined over 50+ phases of development.

When to Use This Skill

  • REQUIRED before making architectural or process changes
  • When checking compliance requirements for work
  • For understanding approval and review requirements
  • When determining quality gates that must pass
  • For escalation and authority questions

Core Governance Framework

The Four Pillars (Canon Charter)

All governance decisions are evaluated against these foundational principles:

  1. Portability 🔄
  2. Cross-environment compatibility
  3. System adaptability
  4. Platform independence
  5. Provider agnosticism

  6. Reproducibility 🔁

  7. Consistent outcomes
  8. Predictable behavior
  9. Deterministic processes
  10. Verifiable results

  11. Safety 🛡️

  12. Risk mitigation
  13. Secure operations
  14. Fail-safe mechanisms
  15. Progressive validation

  16. Research Value 🔬

  17. Scientific methodology
  18. Knowledge advancement
  19. Evidence-based decisions
  20. Learning capture

Authority Structure

Decision Authority Hierarchy

Code Reviewers:
  - Final merge authority
  - Architectural decisions
  - Technical standards
  - Pattern enforcement

PM Agents:
  - Requirements authority
  - Prioritization decisions
  - Workflow coordination
  - Release sign-off

Specialist Roles:
  - Domain expertise
  - Comment-only reviews
  - Advisory input
  - Escalation triggers

Implementers:
  - Cannot merge own PRs
  - Cannot bypass gates
  - Cannot grant exceptions
  - Must follow process

Separation of Concerns

MANDATORY: No agent can approve their own work
- Implementers → Reviewers
- Reviewers → Different reviewer for meta-reviews
- PM defines requirements → Cannot implement
- Security can block any change

Quality Gates

Definition of Done (DoD)

MUST have (blocking):
- ✅ All tests passing
- ✅ Code review approved
- ✅ Documentation updated
- ✅ Journal entry created
- ✅ CI/CD checks green
- ✅ Proper Git/GitHub workflow followed
- Feature branch used (never main/master/develop)
- Draft PR created at task start (not task end)
- No direct commits to default branches
- Project conventions respected (templates, naming)
- Prefer gh CLI over git commands where available
- ✅ PR is appropriately sized (incremental PR strategy)
- <500 lines of actual code (excluding tests/docs)
- <30 files changed
- Provides stand-alone value (can merge without breaking main)
- Can be explained in 2 sentences (clear, focused scope)
- Can be reviewed in <1 hour
- If multi-PR feature: Sequence documented in first PR
- Reference: wolf-workflows/incremental-pr-strategy.md

SHOULD have (strong recommendation):
- 📊 Performance benchmarks met
- 🔒 Security scan clean
- ♿ Accessibility validated
- 📈 Metrics improved

MAY have (optional):
- 🎨 UI/UX review
- 🌍 Internationalization
- 📱 Mobile testing

Two-Tier Test Pipeline

Fast-Lane (5-10 minutes)

Purpose: Rapid iteration and basic validation

Requirements:
- Linting: Max 5 errors allowed
- Unit tests: 60% coverage minimum
- Critical integration tests pass
- Security: 0 critical, ≤5 high vulnerabilities
- Smoke tests: Core services start

Full-Suite (30-60 minutes)

Purpose: Production readiness validation

Requirements:
- E2E tests: 90% success rate
- Performance: Score ≥70/100
- Security: Score ≥80/100
- Cross-platform: Node 18/20/21 compatible
- Migration: Rollback procedures tested

Good/Bad Examples: Definition of Done

Example 1: Feature Pull Request


PR #456: Add user authentication

All MUST-have items complete:
- Tests: 47 tests passing (unit + integration + E2E)
- Fast-Lane: ✅ 8min, 0 linting errors, 85% coverage
- Full-Suite: ✅ 45min, 95% E2E success, perf 75/100, security 85/100
- Review: Approved by @code-reviewer-agent (not self-approved)
- Documentation:
- README.md updated with auth setup instructions
- API.md documents new endpoints
- CHANGELOG.md entry added
- Journal: 2025-11-14-user-authentication.md created
- Problems: OAuth token refresh edge case
- Decisions: Chose JWT over sessions for scalability
- Learnings: Auth middleware testing patterns
- CI: All checks green ✅

SHOULD-have items addressed:
- Performance: Login latency <200ms (target: <300ms) ✅
- Security: OAuth2 threat model documented, scan clean ✅
- Metrics: Login success rate tracking added ✅

Assessment: Meets Definition of Done. ✅ Ready to merge.


PR #457: Fix login bug

Missing MUST-have items:
- Tests: "Tests pass on my machine" (no CI evidence)
- Fast-Lane: Not run
- Full-Suite: Skipped "to save time"
- Review: Self-approved "since it's urgent"
- Documentation: "Will update later"
- Journal: No entry created
- CI: Checks failing (linting errors, 2 test failures)

Governance violations:
- Merged own PR (violates separation of concerns)
- Skipped quality gates "because it's a hotfix"
- No root cause analysis (reliability-fixer archetype requirement)
- No regression test added

Assessment: Does NOT meet Definition of Done. ❌ Should be reverted and reworked.

Why this is wrong:
- Hotfixes still require governance (use expedited review, not no review)
- Self-approval violates authority structure
- Skipping tests means bug might not actually be fixed
- No journal = no learning capture = problem will recur

Example 2: Security Change


PR #789: Implement rate limiting

Security-hardener archetype requirements:
- Threat Model: docs/security/rate-limiting-threats.md
- Attack vectors documented
- Mitigation strategies defined
- Residual risks assessed
- Security Scan: ✅ 0 critical, 2 high (false positives documented)
- Penetration Test: Manual testing results in journal
- Defense-in-Depth: Multiple layers (IP-based, user-based, endpoint-based)

Standard DoD:
- Tests: Rate limit scenarios covered (100% of rate limit logic)
- Review: Approved by @security-agent + @code-reviewer-agent
- Documentation: Rate limit policies documented
- Journal: 2025-11-14-rate-limiting-implementation.md
- CI: All gates green including security gates

ADR created:
- ADR-042-rate-limiting-strategy.md documents algorithm choice

Assessment: Exemplary security change. ✅ All gates passed.


PR #790: Add encryption

Security-hardener failures:
- Threat Model: "Encryption is obviously good" (no actual model)
- Security Scan: Skipped "because I used a well-known library"
- Penetration Test: None performed
- Defense-in-Depth: Single layer only

Code quality issues:
- Using deprecated crypto algorithm (MD5 for hashing)
- Hardcoded encryption keys in code
- No key rotation mechanism
- Error messages leak sensitive info

Missing standard DoD:
- Tests: Only happy path tested
- Review: Only one approval (needs security-agent review)
- Documentation: No encryption policy documented
- Journal: No entry
- ADR: No architecture decision documented

Assessment: Critical security issues. ❌ Must be blocked and reworked.

Why this is dangerous:
- Wrong crypto creates false sense of security
- Hardcoded keys = security theater
- No threat model = don't understand what we're protecting against
- Missing security-agent review = no domain expert validation

Example 3: Refactoring


PR #234: Refactor auth middleware

Maintainability-refactorer archetype requirements:
- Complexity Reduction: Cyclomatic complexity 15 → 6 (documented)
- Test Coverage: Maintained at 85% (no regression)
- Behavior Unchanged: All tests still pass (no behavior changes)

Evidence-based changes:
- Before metrics: 150 LOC, complexity 15, 4 code smells
- After metrics: 95 LOC, complexity 6, 0 code smells
- Performance: No degradation (latency unchanged)

Standard DoD:
- Tests: All existing tests pass + new tests for extracted functions
- Review: Approved by @code-reviewer-agent
- Documentation: Code comments improved, architecture notes updated
- Journal: 2025-11-14-auth-middleware-refactor.md
- Decisions: Extracted 3 functions for clarity
- Learnings: Middleware composition patterns
- CI: All gates green

Assessment: Clean refactoring. ✅ Improves maintainability without risk.


PR #235: Clean up code

Refactoring violations:
- Behavior Changed: Added "small feature" during refactor
- Test Coverage: Dropped from 85% → 60%
- No Metrics: Can't prove complexity improved

Mixed concerns:
- Refactoring + feature + bug fix in one PR
- 847 lines changed across 23 files
- No clear focus or archetype

Missing DoD:
- Tests: 12 tests now failing "will fix in follow-up"
- Review: "Just cleanup, no review needed"
- Documentation: Not updated to reflect changes
- Journal: No entry

Assessment: Dangerous refactoring. ❌ Reject and split into focused PRs.

Why this fails:
- Mixed refactor + feature = impossible to review safely
- Dropping coverage during refactor = introducing bugs
- Behavior change during refactor = violates maintainability-refactorer archetype
- No metrics = can't prove improvement

Process Requirements

Phase Lifecycle (Canonical)

Every phase MUST follow:

  1. Seed Brief 📋
  2. Problem statement
  3. Success criteria
  4. Risk assessment
  5. Resource allocation

  6. Pre-Phase Sweeps 🔍

  7. Dependency check
  8. Conflict resolution
  9. Environment preparation
  10. Baseline metrics

  11. Shard Work

  12. Incremental delivery
  13. Continuous validation
  14. Journal updates
  15. Progress tracking

  16. Close-Out Sweeps

  17. Consolidation
  18. Verification
  19. Documentation
  20. Learning capture

Journal Requirements

MANDATORY for all work:

## Problems
- Issues encountered
- Blockers identified
- Unexpected behaviors

## Decisions
- Choices made
- Trade-offs accepted
- Rationale documented

## Learnings
- Patterns discovered
- Improvements identified
- Knowledge gained

Format: YYYY-MM-DD-<kebab-slug>.md

ADR (Architecture Decision Record)

Required for:
- Architectural changes
- Process modifications
- Tool selections
- Major refactoring

Format:

# ADR-XXX: Title

## Status
[Proposed | Accepted | Deprecated | Superseded]

## Context
[Problem description]

## Decision
[What was decided]

## Consequences
[Trade-offs and impacts]

Compliance Matrix

By Change Type

Change Type Required Approvals Evidence Gates
Bug Fix Code Review Tests, Root Cause CI Pass
Feature PM + Code Review AC Met, Tests DoD Complete
Security Security + Code Review Threat Model, Scan Security Gates
Architecture Architect + Code Review ADR, Impact Analysis Full Suite
Process PM + Architect ADR, Stakeholder Review Governance Check

By Risk Level

Risk Additional Requirements
Low Standard gates
Medium +1 reviewer, extended tests
High Security review, rollback plan
Critical Executive approval, staged rollout

Governance Enforcement

Automated Checks

Pre-commit:
  - Linting
  - Format validation
  - Secrets scanning

CI Pipeline:
  - Test execution
  - Coverage validation
  - Security scanning
  - Performance checks

Pre-merge:
  - Review approval
  - Gate validation
  - Documentation check
  - Journal verification

Manual Reviews

Required human validation:
- Architecture alignment
- Business logic correctness
- Security implications
- UX impact

Change Management

Canon Charter Changes

HIGHEST GOVERNANCE LEVEL
- Research analysis required
- ADR documentation mandatory
- Multi-stakeholder review
- 30-day comment period

Lifecycle Changes

  • ADR required
  • Architect approval
  • Code Reviewer approval
  • Backward compatibility analysis

Policy Updates

  • Impact assessment
  • Migration plan if breaking
  • Communication plan
  • Grace period for adoption

Escalation Paths

Technical Escalation

Developer → Code Reviewer → Architect → CTO

Process Escalation

Agent → PM → Orchestrator → Product Owner

Security Escalation

ANY → Security Agent → CISO → Executive

Emergency Override

Only for production incidents:
1. Document override reason
2. Apply temporary fix
3. Create follow-up ticket
4. Conduct post-mortem

Anti-Patterns (Forbidden)

❌ Process Violations

  • Skipping quality gates
  • Merging own PRs
  • Bypassing security scans
  • Ignoring test failures

❌ Authority Violations

  • Exceeding role boundaries
  • Granting unauthorized exceptions
  • Overriding specialist objections
  • Ignoring escalation requirements

❌ Documentation Violations

  • Missing journal entries
  • No ADR for architecture changes
  • Outdated documentation
  • No evidence for decisions

Governance Metrics

Compliance Indicators

  • Gate pass rate: >95%
  • Review turnaround: <4 hours
  • Journal compliance: 100%
  • ADR coverage: All major changes

Health Metrics

  • CI reliability: >99%
  • Test stability: >95%
  • Security score: >80/100
  • Documentation currency: <7 days

Scripts Available

  • check.js - Validate compliance for current work
  • gates.js - List applicable quality gates
  • escalate.js - Determine escalation path

Integration with Other Skills

  • wolf-principles: Governance implements principles
  • wolf-archetypes: Archetypes follow governance rules
  • wolf-roles: Roles have governance boundaries

Red Flags - STOP

If you catch yourself thinking:

  • "Skipping quality gates to save time" - STOP. Gates exist because skipping them costs MORE time in rework and incidents.
  • "This change is too small for governance" - Wrong. Small changes compound. All work follows governance.
  • "I'll create the journal entry later" - NO. Journal entry is part of Definition of Done. Create it NOW.
  • "Tests are passing locally, CI doesn't matter" - CI is the source of truth. Local != production.
  • "I'm just fixing a typo, no review needed" - ALL changes need review. Separation of concerns is non-negotiable.
  • "We can make an exception this time" - Exceptions become habits. Follow Advisory-First Enforcement (Principle 4).
  • "Documentation can wait until after merge" - NO. Documentation is part of DoD. Must be complete BEFORE merge.
  • "Merging my own PR is faster" - FORBIDDEN. You cannot approve your own work (Authority Structure).

STOP. Use Skill tool to load wolf-governance to check compliance requirements BEFORE proceeding.

After Using This Skill

REQUIRED NEXT STEPS:

Sequential skill chain - DO NOT skip steps
  1. REQUIRED NEXT SKILL: Use wolf-roles to understand role-specific compliance requirements
  2. Why: Governance defines WHAT must be done. Roles define WHO does it and HOW.
  3. Gate: Cannot execute governance without understanding role boundaries
  4. Tool: Use Skill tool to load wolf-roles
  5. Example: pm-agent validates acceptance criteria, coder-agent implements, code-reviewer approves

  6. REQUIRED NEXT SKILL: Use wolf-verification to set up verification checkpoints

  7. Why: Governance gates require verification. Verification skill provides three-layer validation (CoVe, HSP, RAG).
  8. Gate: Cannot claim compliance without verification evidence
  9. When: Always - verification is mandatory for all governance gates
  10. Example: Security gate requires threat model verification + scan validation

  11. REQUIRED BEFORE COMPLETION: Create compliance artifacts

  12. Journal Entry: YYYY-MM-DD-<kebab-slug>.md documenting problems, decisions, learnings
  13. ADR (if applicable): For architectural/process changes
  14. Evidence: Test results, security scans, performance benchmarks
  15. Gate: Cannot merge without complete artifact set

DO NOT PROCEED to merge without completing steps 1-3.

Verification Checklist

Before claiming governance compliance:

  • [ ] Archetype selected (from wolf-archetypes)
  • [ ] Quality gates identified for this work type
  • [ ] Definition of Done requirements understood
  • [ ] Role boundaries confirmed (from wolf-roles)
  • [ ] All MUST-have DoD items completed (tests, review, docs, journal, CI)
  • [ ] SHOULD-have items evaluated (performance, security, a11y, metrics)
  • [ ] Approval from authorized reviewer (not self-approval)
  • [ ] CI/CD checks green (not just local)

Can't check all boxes? Governance compliance incomplete. Return to this skill.

Governance Examples by Change Type

Example 1: Bug Fix (reliability-fixer archetype)

DoD Requirements:
✅ Root cause documented in journal
✅ Regression test added
✅ All tests passing (Fast-Lane + Full-Suite)
✅ Code review approved
✅ CI checks green

Gates:
- Fast-Lane: <10 min, linting ≤5 errors, 60% coverage
- Full-Suite: 90% E2E success, rollback tested
- Review: Code reviewer approval (not self)

Example 2: Feature (product-implementer archetype)

DoD Requirements:
✅ Acceptance criteria met (PM validation)
✅ Tests pass (unit + integration + E2E)
✅ Documentation updated (README, API docs)
✅ Journal entry created
✅ Performance benchmarks met

Gates:
- PM: Acceptance criteria validation
- Code Review: Technical quality + tests
- Security: Scan clean (if data handling involved)
- CI: Fast-Lane + Full-Suite green

Example 3: Security Change (security-hardener archetype)

DoD Requirements:
✅ Threat model documented
✅ Security scan clean
✅ Penetration test passed
✅ Defense-in-depth applied
✅ Monitoring/alerting added

Gates:
- Security Review: Threat model approved
- Security Scan: 0 critical, ≤5 high vulns
- Code Review: Implementation quality
- CI: Security gates + standard gates
- ADR: If architectural security decision

Emergency Override Procedure

ONLY for production incidents:

1. Document override reason immediately
2. Apply minimum necessary temporary fix
3. Create follow-up ticket for proper fix
4. Conduct post-mortem within 48 hours
5. Update governance if process failed

Override does NOT mean skip governance. It means parallel governance with expedited review.


Source: docs/governance/, Canon Charter, ADRs
Last Updated: 2025-11-14
Phase: Superpowers Skill-Chaining Enhancement v2.0.0*

# Supported AI Coding Agents

This skill is compatible with the SKILL.md standard and works with all major AI coding agents:

Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.