PSD Claude Coding System

Peninsula School District's comprehensive Claude Code plugin for AI-assisted software development.

Overview

One unified plugin combining battle-tested development workflows with memory-based learning and knowledge compounding.

Version: 1.25.1 Status: Production-Ready Workflows + Memory-Based Learning

Quick Start

# Install the marketplace
/plugin marketplace add psd401/psd-claude-coding-system

# Install the plugin
/plugin install psd-claude-coding-system

# Start using immediately
/work 347              # Implement an issue
/test                  # Run tests
/evolve               # Auto-evolve the plugin

What's New in v1.25.1

Iterative /review-pr — Multi-Round PR Feedback Handling

Key Changes

Incremental detection — /review-pr now detects previous runs via PR comment markers and only processes NEW feedback since the last round
Round tracking — Each run embeds  in the summary comment, enabling stateless cross-session tracking
Smart agent dispatch — Always-on structural review agents (architecture-strategist, code-simplicity-reviewer, pattern-recognition-specialist) run on Round 1 only; rounds 2+ focus on addressing reviewer feedback
Early exit — If no new feedback exists since last round, exits with "PR up to date" message
--full flag — Force a complete re-review: /review-pr 123 --full

How It Works

PSD Claude Coding System - Complete Workflow

The plugin provides a complete development lifecycle:

Planning - Create issues (/issue) or break down big ideas into epics (/product-manager), design architecture (/architect)
Implementation - Work on issues (/work 347) with automatic agent assistance
Validation - Run tests (/test), handle PR feedback (/review-pr)
Completion - Clean up (/clean-branch), evolve the plugin (/evolve)

Workflow Commands

Command	Description	Example
`/work`	Implement solutions with auto reviews + learning capture	`/work 347`
`/lfg`	Autonomous end-to-end: implement → test → review → fix → learn	`/lfg 347`
`/architect`	System architecture design	`/architect "caching system"`
`/test`	Comprehensive testing with self-healing + learning capture	`/test auth`
`/review-pr`	Iterative PR feedback (rounds 2+ process only new comments)	`/review-pr 123`
`/issue`	AI-validated GitHub issues	`/issue "add caching"`
`/product-manager`	Product specs to sub-issues	`/product-manager "dashboard"`
`/security-audit`	Security analysis	`/security-audit`
`/scope`	Scope classification + tiered planning	`/scope "add caching"`
`/evolve`	Auto-evolve: analyze learnings, check releases, compare plugins	`/evolve`
`/clean-branch`	Post-merge cleanup	`/clean-branch`
`/triage`	FreshService ticket triage	`/triage 12345`

AI Agents (42 total)

Review Specialists (14 agents)

security-analyst · security-analyst-specialist · deployment-verification-agent · data-migration-expert · agent-native-reviewer · architecture-strategist · code-simplicity-reviewer · pattern-recognition-specialist · schema-drift-detector · data-integrity-guardian · typescript-reviewer · python-reviewer · swift-reviewer · sql-reviewer

Domain Specialists (7 agents)

backend-specialist · frontend-specialist · database-specialist · llm-specialist · ux-specialist · architect-specialist · shell-devops-specialist

Quality (3 agents)

test-specialist · performance-optimizer · documentation-writer

Research (6 agents)

learnings-researcher · spec-flow-analyzer · best-practices-researcher · framework-docs-researcher · git-history-analyzer · repo-research-analyst

Workflow (4 agents)

bug-reproduction-validator · work-researcher · work-validator · learning-writer

Meta & Validation (6 agents)

meta-reviewer · plan-validator · document-validator · configuration-validator · breaking-change-validator · telemetry-data-specialist

External AI (2 agents)

gpt-5-codex (GPT-5.3-Codex) · gemini-3-pro (Gemini 3.1 Pro)

Knowledge Compounding System

The plugin includes a hybrid knowledge capture system:

Project Learnings (`./docs/learnings/`)

Store project-specific patterns and solutions (local only, gitignored)
Automatically searched before implementation via learnings-researcher
Automatically captured by /work, /test, /review-pr, /lfg via learning-writer agent
Auto-deleted after 90 days by /evolve TTL cleanup

Plugin Patterns (`docs/patterns/`)

Universal patterns shared across all projects
Contributed via /evolve (auto-detects uncontributed universal learnings)
Available to all plugin users after merge

Language-Specific Reviews

The plugin automatically detects languages in changed files and invokes appropriate reviewers:

Language	Extensions	Reviewer
TypeScript	`.ts`, `.tsx`, `.js`, `.jsx`	`typescript-reviewer`
Python	`.py`	`python-reviewer`
Swift	`.swift`	`swift-reviewer`
SQL	`.sql`, `migration`	`sql-reviewer`

Dual-Phase Review

Light Mode (pre-PR in /work): Quick checks, critical issues only
Full Mode (post-PR in /review-pr): Comprehensive deep analysis

Installation

From GitHub (Recommended)

/plugin marketplace add psd401/psd-claude-coding-system
/plugin install psd-claude-coding-system

Verify Installation

/plugin list
# Should show: psd-claude-coding-system (v1.25.1)

# Test a command
/evolve

Usage Examples

Basic Workflow

# Work on an issue
/work 347

# Run tests
/test

# Create PR and handle feedback
/review-pr 123

# Clean up after merge
/clean-branch

# Evolve the plugin (auto-picks highest-value action)
/evolve

Architecture

plugins/psd-claude-coding-system/
├── skills/                    # 12 user-invocable skills
│   ├── work/SKILL.md          # Main implementation workflow
│   ├── lfg/SKILL.md           # Autonomous end-to-end workflow
│   ├── evolve/SKILL.md        # Auto-evolve (learnings, releases, comparison)
│   └── ...                    # Other workflow skills
├── agents/                    # 42 specialized agents
│   ├── review/                # 14 code review specialists
│   ├── domain/                # 7 domain experts
│   ├── quality/               # 3 quality assurance
│   ├── research/              # 6 research agents
│   ├── workflow/              # 4 workflow agents
│   ├── external/              # 2 external AI providers
│   ├── meta/                  # 1 meta-reviewer
│   └── validation/            # 5 validators
├── docs/
│   ├── learnings/             # Project learnings (auto-captured)
│   └── patterns/              # Universal patterns
├── scripts/
│   ├── post-edit-validate.sh  # PostToolUse syntax validation
│   └── language-detector.sh   # Language detection utility
└── hooks/
    └── hooks.json             # PostToolUse validation hook

Compound Engineering Principles

Every interaction creates improvement opportunities:

Every bug → prevention system
Every manual process → automation candidate
Every solution → template for similar problems
Every workflow → captured learning for future sessions

Use /evolve to analyze accumulated learnings and improve the plugin. /work, /test, /review-pr, and /lfg always dispatch the learning-writer agent automatically (it handles deduplication).

Documentation

Plugin README - Detailed plugin docs
CLAUDE.md - Technical reference

Support

Author: Kris Hagel (hagelk@psd401.net)
Organization: Peninsula School District
Repository: psd401/psd-claude-coding-system
Issues: https://github.com/psd401/psd-claude-coding-system/issues

Acknowledgments

Every's Compound Engineering Plugin - Inspiration for v1.15.1 knowledge compounding and language-specific review patterns
Built with Claude Code by Anthropic

License

MIT License - see LICENSE for details

Peninsula School District - Innovating education through technology

Name	meta-evolve
Description	Evolve agent prompts using genetic algorithms and historical performance data

meta-evolve

SKILL.md