workflow-enforcement
Protocol-based workflow enforcement with validation dependencies and anti-bypass protection
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.
| name | workflow-enforcement |
|---|---|
| description | Protocol-based workflow enforcement with validation dependencies and anti-bypass protection |
| category | architecture |
| 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_allandform_selection_auditcan run together - Sequential:
functional_testrequires 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-modeflag
MCP Integration
- SessionStart hook auto-starts MCP daemons
mcp-health-check.pyverifies daemon status- Optional
mcp_initializedvalidation for MCP-dependent workflows
References
references/state-machine-patterns.md- State machine designreferences/gate-design.md- Gate placement and designreferences/phase-transition.md- Phase sequencing rulesreferences/exit-code-guide.md- Exit code classificationreferences/protocol-design.md- 6 workflow types detailed