AI-powered development workflow automation for Claude Code. Phase-based planning, implementation orchestration, preflight code quality checks, and ship-it workflow for efficient sub-agent execution.
Provides tools for managing complex development workflows including breaking features into context-efficient phases, orchestrating multi-phase implementations via sub-agents, running comprehensive code quality checks, and shipping code with a single command.
- Phase Planning: Break large features into properly-sized phases (30-50k tokens each) optimized for sub-agent execution
- Implementation Orchestration: Analyze dependencies and execute phases with optimal parallel/sequential strategies
- Preflight Checks: Auto-detect and run type checking, linting, formatting, security scanning, and tests across multiple ecosystems
- Ship It: Run preflight checks, commit, push, and create a PR in one streamlined flow
Create structured implementation plans from feature descriptions, optimized for context-efficient sub-agent execution.
What it does:
- Asks clarifying questions to understand scope and requirements
- Breaks features into right-sized phases (30-50k tokens each)
- Uses whole number phases only (no sub-phases like 1.1, 1.2)
- Defines clear acceptance criteria per phase
- Maps dependencies between phases
- Recommends execution strategy (parallel/sequential/mixed)
- Outputs structured markdown to
docs/plans/
Usage:
/workflow-plan-phases "Build a user authentication system with OAuth, MFA, and session management"
/workflow-plan-phases --output=docs/plans/auth-system.md "Build authentication system..."
Before (manual planning):
# Manually break down feature into tasks
# Guess at task sizes and dependencies
# Hope sub-agents don't run out of context
# Lose track of what depends on what
After (with ai-workflow plugin):
/workflow-plan-phases "Build user authentication"
# Answers clarifying questions
# AI creates optimally-sized phases
# Dependencies clearly mapped
# Execution strategy recommended
# Plan saved to docs/plans/
Orchestrate multi-phase implementation from a plan document using intelligent parallel/sequential execution.
What it does:
- Parses plan documents and extracts phase definitions
- Analyzes dependencies (explicit and implicit)
- Determines optimal execution strategy
- Presents execution plan for user confirmation
- Executes via Task() sub-agents with coordination directory
- Handles failures gracefully (skips dependent phases, continues independent ones)
- Aggregates results and provides comprehensive summary
Usage:
/workflow-implement-phases @docs/plans/user-authentication.md
/workflow-implement-phases @docs/plans/feature.md --phases=1,2,3
/workflow-implement-phases @docs/plans/feature.md --strategy=parallel
Execution Strategies:
| Strategy | When Used | Benefits |
|---|---|---|
| Parallel | No dependencies between phases | Fastest execution, maximum context isolation |
| Sequential | Linear dependency chain | Safest execution, full context from prior phases |
| Mixed | Some independent, some dependent | Best of both worlds |
Run comprehensive code quality checks before commits, PRs, or deployments.
What it does:
- Auto-detects configured quality tools across ecosystems
- Runs checks in optimal order: format -> typecheck -> lint -> security -> tests
- Detects security tools: pnpm/npm/yarn audit, eslint-plugin-security, Semgrep
- Universal Semgrep detection via config files, CI workflows, README docs, or Docker fallback
- Reports results with clear pass/fail/warning indicators
- Offers interactive fix mode (or use
--fixfor automatic) - Respects existing project scripts (uses
npm run lintover raweslint)
Usage:
/workflow-preflight # Interactive mode (default)
/workflow-preflight --fix # Auto-fix all fixable issues
/workflow-preflight --check-only # Report only, no fixes
/workflow-preflight --verbose # Show detailed output
Supported Ecosystems:
| Ecosystem | Type Check | Lint | Format | Security | Test |
|---|---|---|---|---|---|
| Node.js/TypeScript | tsc | ESLint, Biome | Prettier | pnpm/npm/yarn audit, eslint-plugin-security, Semgrep | Jest, Vitest |
| Python | MyPy | Ruff | Black, Ruff | pip-audit, safety, Semgrep | Pytest |
| .NET | dotnet build | Analyzers | dotnet format | Semgrep | dotnet test |
| Go | go build | golangci-lint | gofmt | Semgrep | go test |
| Rust | cargo check | Clippy | cargo fmt | cargo audit, Semgrep | cargo test |
Before (manual checks):
# Remember which tools are configured
# Run each check manually in the right order
# Parse output to find issues
# Decide whether to fix or not
# Run checks again to verify
After (with ai-workflow plugin):
/workflow-preflight
# AI discovers configured tools
# Runs all checks in optimal order
# Shows clear summary
# Offers to fix issues
# Re-verifies after fixes
/plugin install ai-workflow@claude-code-plugins-dev
# Plan a new feature
/workflow-plan-phases "Add user profile editing with avatar upload"
# Review the generated plan
# Located at: docs/plans/user-profile-editing.md
# Implement the phases
/workflow-implement-phases @docs/plans/user-profile-editing.md
# Run preflight checks before committing
/workflow-preflight
# Or ship everything in one go (preflight + commit + push + PR)
/workflow-ship
User Description
|
v
Clarifying Questions <-- Never skipped
|
v
Requirements Analysis
|
v
Phase Sizing (30-50k tokens each)
|
v
Dependency Mapping
|
v
Execution Strategy
|
v
Plan Document (docs/plans/)
Plan Document
|
v
Parse Phases & Extract Specs
|
v
Analyze Dependencies
|
v
Build Execution Graph
|
v
Present Plan to User
|
v
Execute via Sub-Agents
|
v
Aggregate Results
|
v
Summary Report
Discovery Phase
|
v
Detect Project Type(s)
|
v
Find Configured Tools (including security scanners)
|
v
Run Checks (format -> type -> lint -> security -> test)
|
v
Present Results
|
v
Fix Prompt (if issues found)
|
v
Verify Fixes
- Large features that would exceed a single context window
- Complex implementations with multiple dependencies
- Team projects where different phases could be assigned to different people
- Features requiring careful dependency management
| Phase Size | Tokens | Files | Recommendation |
|---|---|---|---|
| Too Small | <15k | 1 | Combine with related work |
| Right-Sized | 30-50k | 2-5 | Ideal for sub-agents |
| Too Large | >60k | 6+ | Split by layer, entity, or concern |
- Run
/workflow-preflightbefore every commit - Use
--fixin CI for automated formatting - Use
--check-onlyin CI for validation gates - Align local checks with CI configuration
Per feature implementation:
- Manual planning: ~1-2 hours (for complex features)
- With plan-phases: ~10-15 minutes
Per preflight check:
- Manual checks: ~5-10 minutes (remembering commands, running each tool)
- With preflight: ~1-2 minutes (automated discovery and execution)
Estimated monthly savings:
- Phase planning: Save ~4-8 hours
- Preflight checks: Save ~2-4 hours
- Total: Save ~6-12 hours/month
Plus improved code quality, fewer context overflows, and more efficient sub-agent execution.
- Name: AI-Workflow
- Version: 1.3.0
- Type: Development Workflow Automation
- Features:
- Skills:
/workflow-plan-phases,/workflow-implement-phases,/workflow-preflight,/workflow-ship - Security scanning: pnpm/npm/yarn audit, eslint-plugin-security, Semgrep (CLI or Docker)
- Skills:
- License: MIT
- Author: Charles Jones
Found a bug or have a suggestion? Open an issue or submit a pull request!
MIT License - See LICENSE file for details.
Built with love for the Claude Code community