workflow-enforcement

chkim-su's avatarfrom chkim-su

Protocol-based workflow enforcement with validation dependencies and anti-bypass protection

0stars🔀0forks📁View on GitHub🕐Updated Jan 7, 2026

When & Why to Use This Skill

This Claude skill implements a robust protocol-based workflow enforcement system designed to ensure process integrity through strict validation dependencies and anti-bypass protection. By utilizing dependency graphs and state-machine patterns, it prevents agents or users from skipping critical quality gates, making it an essential tool for maintaining high standards in automated software development and agentic operations.

Use Cases

  • Automated Skill and Agent Development: Enforce a multi-stage validation pipeline (quality audits, functional testing, and form audits) that must be completed before a new AI skill or agent is finalized.
  • Secure Marketplace Publishing: Implement blocking validation gates for plugin deployments, ensuring that content quality audits and security checks cannot be bypassed during the publishing process.
  • CI/CD Pipeline Integrity: Integrate with development hooks to trigger mandatory validation scripts, using exit code policies to allow, warn, or block workflow transitions based on real-time test results.
  • Complex Task Orchestration: Manage intricate workflows with parallel and sequential dependencies, ensuring that prerequisite tasks like 'validate_all' are successfully completed before proceeding to functional testing.
nameworkflow-enforcement
descriptionProtocol-based workflow enforcement with validation dependencies and anti-bypass protection
categoryarchitecture
tools[]

Workflow Enforcement Patterns

Protocol-based workflow enforcement with dependency graphs and anti-bypass protection.

Core Concept

validate_all ──┬──> content_quality_audit
               │
form_audit ────┼──> functional_test ──> plugin_test
               │
               └──> (MCP workflows)

Workflow Types

Type Description Required Validations
skill_creation Creating new skills validate_all, form_audit, functional_test
agent_creation Creating new agents validate_all, form_audit
command_creation Creating commands validate_all
plugin_publish Marketplace deployment ALL validations + content_quality (blocking)
quick_fix Simple error fixes validate_all only
analyze_only Read-only analysis validate_all

Validation Dependencies

validate_all (no deps) ──┬──> content_quality_audit
                         │
form_selection_audit ────┼──> functional_test ──> plugin_test
(no deps)                │
  • Parallel: validate_all and form_selection_audit can run together
  • Sequential: functional_test requires both to pass first

Anti-Bypass Protection

Agent-required validations cannot be manually passed:

# This will FAIL (manual bypass attempt)
python3 forge-state.py mark-validation form_selection_audit passed

# This will SUCCEED (triggered by hook after agent completes)
python3 forge-state.py mark-validation form_selection_audit passed --from-hook

Protected validations: form_selection_audit, functional_test, plugin_test

Quick Start

# Initialize workflow
python3 scripts/forge-state.py init skill_creation

# Check status
python3 scripts/forge-state.py status

# Check dependencies before validation
python3 scripts/forge-state.py check-deps functional_test

# Mark validation (via hook only for protected ones)
python3 scripts/forge-state.py mark-validation validate_all passed

# Verify protocol completion
python3 scripts/forge-state.py verify-protocol

Exit Code Policy

Exit Code Meaning Hook Behavior
exit(0) ALLOW Proceed
exit(1) WARN May block
exit(2) BLOCK Always blocks in PreToolUse

Content Quality Validation

W037 (Korean) and W038 (emoji) warnings:

  • Normal mode: Warning only (exit 0)
  • Publish mode: Blocking (exit 2) via --publish-mode flag

MCP Integration

  • SessionStart hook auto-starts MCP daemons
  • mcp-health-check.py verifies daemon status
  • Optional mcp_initialized validation for MCP-dependent workflows

References

  • references/state-machine-patterns.md - State machine design
  • references/gate-design.md - Gate placement and design
  • references/phase-transition.md - Phase sequencing rules
  • references/exit-code-guide.md - Exit code classification
  • references/protocol-design.md - 6 workflow types detailed
workflow-enforcement – AI Agent Skills | Claude Skills