Apex Spec System
Version: 2.0.4-beta
A Claude Code plugin providing a specification-driven workflow system for AI-assisted development. Think Github Spec Kit (our source inspiration) simplified.
NEW - Apex Spec System "Open"
Open Skill (https://agentskills.io/home) version of this plugin: https://github.com/moshehbenavraham/apex-spec-system-open
Overview
The Apex Spec System breaks large projects into manageable, well-scoped implementation sessions that fit within AI context windows and human attention spans.
Philosophy: 1 session = 1 spec = 2-4 hours (12-25 tasks) = safe context window of AI
Installation
# Install from local directory
claude --plugin-dir /path/to/apex-spec-system
# Or copy to your plugins directory
cp -r apex-spec-system ~/.claude/plugins/
Requirements
| Dependency | Required | Install |
|---|---|---|
| jq | Yes | apt install jq or brew install jq |
| git | Optional | Usually pre-installed |
The scripts use jq for JSON parsing. Verify with: bash scripts/check-prereqs.sh --env
Quick Start
Install the plugin (see above)
Initialize in your project:
/initspec OR /apex-spec:initspecThis creates the spec system structure in your project.
Optional but recommended: Customize
.spec_system/CONVENTIONS.mdwith your project's coding standards (naming, structure, error handling, testing philosophy, etc.)/createprd OR /apex-spec:createprd OR Manually fill out .spec_system/PRD/PRD.mdOptional: Turn argument or file path into a technical PRD for development. Example: /createprd "a habit trackker app" /createprd @docs/requirements.md
/createuxprd OR /apex-spec:createuxprdOptional: Turn design docs into a UX PRD companion document. Example: /createuxprd "social app with feed, profiles, and messaging" /createuxprd @docs/design-spec.md
/phasebuild OR /apex-spec:phasebuildThis will set up the initial Phase and Sessions for that initial Phase
Run the session workflow and repeat until all sessions inside the Phase are completed, thus completing the Phase:
/plansession OR /apex-spec:plansession # Analyze, spec, and generate task checklist /implement OR /apex-spec:implement # Start implementation /validate OR /apex-spec:validate # Verify completeness, security & compliance /updateprd OR /apex-spec:updateprd # Mark complete, update systemBetween Phases
/audit OR /apex-spec:audit # Local dev tooling (formatter, linter, types, tests, observability, hooks) /pipeline OR /apex-spec:pipeline # CI/CD workflows (quality, build, security, integration, ops) /infra OR /apex-spec:infra # Production infrastructure (health, security, backup, deploy) /carryforward OR /apex-spec:carryforward # Lessons learned, security/compliance records /documents OR /apex-spec:documents # Create, maintain project documentation -- Optional but HIGHLY recommended, do manual testing and separate LLM Audit HERE -- /phasebuild OR /apex-spec:phasebuild # Set up next Phase and Phase's sessionsRepeat until all phases complete!
Monorepo Quick Start
The system auto-detects monorepo structures. No special configuration needed.
Brownfield (existing monorepo with code):
/initspec
The system detects workspace configs (pnpm, npm workspaces, turbo, nx, cargo, go, lerna), shows detected packages, and asks you to confirm.
Greenfield (new project, PRD describes multiple services):
/initspec # Sets monorepo: null (unknown)
/createprd # Detects multi-package signals in PRD, prompts to confirm
Per-package sessions:
User: /plansession
"Plan a session for apps/web"
# Claude scopes spec.md and tasks.md to apps/web
# File paths use full repo-root-relative paths (apps/web/src/auth.ts)
Sessions interleave across packages within a phase. A phase completes when all its sessions are done, regardless of which packages they target.
What stays the same:
- Single
.spec_system/at the repo root - Same session ID format (
phaseNN-sessionNN-name) - Same 13-command workflow
- Single-repo projects see zero change
Features
- 13-Command Workflow: Structured process from initialization to completion
- Session Scoping: Keep work manageable with 12-25 tasks per session
- Progress Tracking: State file and checklists track progress
- Validation Gates: Verify completeness, security & GDPR compliance before marking done
- Coding Conventions: Customizable standards enforced during implementation and validation
- Auto-Activating Skill: Provides workflow guidance automatically
- Dev Tooling: Regular code quality audits
- Documentation Maintenance: Keep project documentation up to date
- ASCII Enforcement: Avoid encoding issues that break code generation
- Monorepo Support: Auto-detects workspace structures, per-package session scoping
Plugin Components
Commands (13 total)
| Command | Purpose |
|---|---|
/initspec |
Initialize spec system in current project |
/createprd |
Generate master PRD from requirements document |
/createuxprd |
Generate UX PRD from design documents |
/plansession |
Analyze project, create spec and task checklist |
/implement |
AI-led task-by-task implementation |
/validate |
Verify session completeness |
/updateprd |
Mark session complete, sync documentation |
/audit |
Local dev tooling (formatter, linter, types, tests, observability, hooks) |
/pipeline |
CI/CD workflows (quality, build, security, integration, ops) |
/infra |
Production infrastructure (health, security, backup, deploy) |
/documents |
Audit and update project documentation |
/carryforward |
Extract lessons learned, maintain security/compliance record between phases |
/phasebuild |
Create structure for new phase |
Utility Commands
Standalone helpers that run outside the session workflow -- safe at any time.
| Command | Purpose |
|---|---|
/sculpt-ui |
Guide AI-led creation of distinctive, production-grade frontend interfaces |
/copush |
Pull, version-bump, commit all changes, and push to origin |
/dockcleanbuild |
Clean Docker environment and rebuild all images and containers from scratch |
/dockbuild |
Quick Docker Compose build and start with full output |
/up2imp |
Audit upstream changes and curate an optimally ordered implementation list |
/pullndoc |
Git pull an upstream repo and document every imported change |
/qimpl |
Context-aware autonomous implementation session driven by a work file |
/qfrontdev |
Autonomous frontend implementation session with designer-level quality standards |
/qbackenddev |
Autonomous backend/infrastructure development session driven by a work file |
As utility commands are added, they will be listed here. See Utility Commands Reference for conventions and details.
Skill
The spec-workflow skill auto-activates when:
- Working in projects with
.spec_system/directory - User mentions spec system concepts
- User asks about session workflow
Bundled Resources
- scripts/: Bash utilities for project analysis
Project Structure
After running /initspec, your project will have:
your-project/
|-- .spec_system/ # All spec system files
| |-- state.json # Project state tracking
| |-- CONSIDERATIONS.md # Institutional memory (lessons learned)
| |-- SECURITY-COMPLIANCE.md # Security posture & GDPR compliance
| |-- CONVENTIONS.md # Project coding standards and conventions
| |-- PRD/ # Product requirements
| | |-- PRD.md # Master PRD
| | \-- phase_00/ # Phase definitions
| |-- specs/ # Implementation specs
| | \-- phaseNN-sessionNN-name/
| | |-- spec.md
| | |-- tasks.md
| | |-- implementation-notes.md
| | |-- security-compliance.md
| | \-- validation.md
| \-- archive/ # Completed work
\-- (your project source files)
For monorepo projects, the same .spec_system/ sits at the repo root. Sessions reference their target package in metadata (spec.md header), not in directory names.
Session Scope
- Maximum 25 tasks per session
- Maximum 4 hours estimated time
- Single clear objective
- Ideal: 12-25 tasks (sweet spot: 20)
ASCII Encoding (Non-Negotiable)
All files must use ASCII-only characters (0-127). No Unicode, emoji, or smart quotes.
Video Tutorial
Watch on YouTube - Installation and workflow walkthrough
Documentation
- Usage Guidance - When to use, workflow modes, team patterns, monorepo guidance
- Production Walkthrough - Real-world examples (single-repo and monorepo)
Apex Infinite CLI
The apex-infinite-cli/ directory contains a standalone Python CLI that runs the entire Apex Spec System workflow autonomously in an infinite loop. A manager LLM decides which command to run next, executes Claude Code as a subprocess, logs the interaction to SQLite, and repeats -- no human intervention required unless the manager asks for help.
Supports Grok, OpenAI, and Ollama as the manager LLM. See apex-infinite-cli/README-apex-infinite-cli.md for setup and usage.
License
MIT License - Use freely in your projects.