My Claude Code setup for software engineering, SOC operations, and product management - 51 skills, 34 commands, 7 custom MCP servers, 14 plugins, custom hooks, and multi-agent orchestration.
All PII, API keys, passwords, and server URLs have been stripped. This is a reference/inspiration repo, not a drop-in config.
Custom Python MCP servers plus third-party servers for external tool access:
| Server | Package / Runtime | Purpose |
|---|---|---|
| brave-search | @brave/brave-search-mcp-server (npx) |
Web search via Brave API |
| puppeteer | @modelcontextprotocol/server-puppeteer (npx) |
Headless browser automation |
| sequential-thinking | @modelcontextprotocol/server-sequential-thinking (npx) |
Chain-of-thought reasoning |
| opensearch | Custom Python server | Query OpenSearch clusters (logs, errors, traces) |
| atlassian | @anthropic/mcp-server-atlassian (npx) |
Jira ticket management |
| prometheus | Custom Python server | Prometheus metrics queries and alerting |
| notion | Custom Python server (uv + httpx) | Notion API via cookie auth |
| soc-investigations | Custom Python server (uv + psycopg2) | Example SOC investigation queries against a Postgres DB |
| soc-infrastructure | Custom Python server (uv + psycopg2) | Example infrastructure health checks (pods, OOMs, disk, queues) |
| soc-integrations | Custom Python server (uv + psycopg2) | Example integration health and error monitoring |
| gmail | @gongrzhe/server-gmail-autoauth-mcp (npx) |
Gmail read/send/draft with OAuth |
| computer-use | Anthropic Computer Use MCP | Native desktop app automation |
| chrome-devtools | Chrome DevTools MCP | Browser DOM interaction |
| macos-automator | macOS Automator MCP | AppleScript/JXA execution |
| exa | Exa search MCP | AI-native web search |
mcp-servers/
opensearch-direct/ # Direct OpenSearch queries (Python)
opensearch-dashboards/ # OpenSearch dashboards proxy (Python)
prometheus/ # PromQL queries (Python)
notion/ # Notion internal API (Python)
soc-investigations/ # Alert & investigation queries
soc-infrastructure/ # Pod health, OOMs, disk IO
soc-integrations/ # Integration status & errors
shared/ # Shared utilities (DB, kubectl, Teams webhook)
14 installed plugins (12 active):
| Plugin | Marketplace | Status | Purpose |
|---|---|---|---|
| plugin-dev | claude-code-plugins | Active | Plugin development toolkit (create skills, agents, hooks) |
| canvas | claude-canvas | Active | Terminal TUI canvases (calendars, documents, flights) |
| claude-delegator | jarrodwatts-claude-delegator | Active | Delegate tasks to Codex/other agents |
| pm-toolkit | pm-skills | Active | NDA drafts, resume review, proofreading, privacy policies |
| pm-product-strategy | pm-skills | Active | SWOT, PESTLE, Porter's 5 Forces, pricing, business models |
| pm-product-discovery | pm-skills | Active | Feature prioritization, experiments, interview scripts |
| pm-market-research | pm-skills | Active | Competitor analysis, personas, journey maps, market sizing |
| pm-data-analytics | pm-skills | Active | Cohort analysis, A/B testing, SQL generation |
| pm-marketing-growth | pm-skills | Active | North Star metrics, positioning, product naming |
| pm-go-to-market | pm-skills | Active | GTM strategy, growth loops, battlecards, ICP |
| pm-execution | pm-skills | Active | PRDs, OKRs, sprints, user stories, stakeholder maps |
| ralph-loop | claude-plugins-official | Disabled | Recurring task loops |
| ralph-wiggum | claude-code-plugins | Disabled | Companion blob |
| frontend-design | claude-code-plugins | Disabled | Frontend design assistance |
- pm-skills from
phuryn/pm-skills(GitHub) - Full PM toolkit suite
| Skill | Description |
|---|---|
/ship |
Full ship workflow - merge base, tests, review, version bump, changelog, PR |
/review |
Pre-landing PR review (SQL safety, LLM trust boundaries, side effects) |
/review-pr |
Review PR and leave comments |
/review-pr-dual |
Dual-model validation (Codex + Opus) with cross-validated findings |
/review-pr-codex |
Review PR with OpenAI Codex |
/check-pr |
Check PR for unresolved comments, failing checks |
/understand-pr |
Navigate large PRs systematically - file by file |
/fix-ci |
Fix CI/CD pipeline issues |
/fix-migrations |
Fix Django migration conflicts after merging main |
/investigate |
Systematic debugging with root cause investigation (4 phases) |
/git-pr-bug-attribution |
Determine whether a bug was introduced by a specific PR |
/simplify |
Review changed code for reuse, quality, efficiency |
/greploop |
Loop Greptile review until zero comments |
/document-release |
Post-ship documentation updates |
/land-and-deploy |
Merge PR, wait for CI, verify production health |
/deploy-staging |
Deploy to staging environment |
/deploy-preprod |
Deploy to preprod environment |
/setup-deploy |
Configure deployment platform settings |
/prod-verify |
Verify PR code changes against prod data |
/codex |
OpenAI Codex CLI - review, challenge, delegate modes |
/codex-swarm |
Spawn parallel Codex agents for subtasks |
/orchestrate |
Orchestrate multiple Claude/Codex agents in tmux panes |
/swarm-review |
6 parallel agents - Docker, frontend, Codex, Playwright, Opus, live |
/retro |
Weekly engineering retrospective with trend tracking |
/weekly-report |
Generate weekly report from GitHub activity |
| Skill | Description |
|---|---|
/browse |
Fast headless browser for QA and site dogfooding |
/gstack |
Headless browser CLI (Playwright-based) for testing |
/gstack-upgrade |
Upgrade gstack to latest version |
/connect-chrome |
Launch real Chrome controlled by gstack with Side Panel |
/qa |
Systematically QA test web app and fix bugs found |
/qa-only |
Report-only QA testing with health score and screenshots |
/design-review |
Designer's eye QA - visual inconsistency, spacing, hierarchy |
/canary |
Post-deploy canary monitoring for errors and regressions |
/benchmark |
Performance regression detection with browse daemon |
/teardown |
Reverse-engineer web app architecture from frontend |
/setup-browser-cookies |
Import cookies from real browser into headless session |
/agent-browser |
Browser automation via agent-browser CLI |
/electron |
Automate Electron desktop apps via CDP |
| Skill | Description |
|---|---|
/soc |
SOC Agent troubleshooting - investigations, infra, integrations |
/status |
Error status report with RCA, stacktraces, affected tenants |
/alert-errors |
Check OpenSearch for errors, create Zenduty incidents |
/cso |
Chief Security Officer mode - infrastructure-first security audit |
/opensearch |
Query OpenSearch logs with flexible filters |
/opensearch-traceback-retrieval |
Retrieve full Python stack traces from structured logs |
/prometheus |
Query Prometheus metrics, health checks, PromQL |
/langfuse |
Langfuse LLM observability insights |
/rancher |
Rancher API - clusters, kubeconfig, nodes, workloads |
| Skill | Description |
|---|---|
/notion |
Interact with Notion's internal API |
/jira |
Create/manage Jira tickets |
/n8n |
Interact with n8n workflows |
/tmux |
Manage Claude Code sessions in tmux |
/find-chat |
Search Claude Code chat history and open matching session |
/deep-research |
Deep research with Graph of Thoughts - parallel exploration |
/youtube-to-ebook |
Transform YouTube videos into formatted ebook articles |
/learn |
Learning skill |
| Skill | Description |
|---|---|
/devils-advocate |
Adversarial argument loop - steelman then demolish |
/frontend-slides |
Create animation-rich HTML presentations |
| Skill | Description |
|---|---|
/plan-ceo-review |
CEO/founder-mode plan review - 10-star product thinking |
/plan-eng-review |
Eng manager plan review - architecture, data flow, edge cases |
/plan-design-review |
Designer's eye plan review - rates each dimension 0-10 |
/autoplan |
Auto-review pipeline - runs CEO, design, eng reviews sequentially |
/planning-with-files |
Manus-style file-based planning with task_plan.md |
/design-consultation |
Full design system proposal (aesthetic, typography, colors) |
/product-management |
PM mode for non-developers - bridge design and engineering |
/vibe-coding |
Vibe coding best practices for rapid prototyping |
| Skill | Description |
|---|---|
/install-mcp |
Install and configure MCP servers |
/create-command |
Create Claude Code slash commands |
/freeze / /unfreeze |
Restrict/unrestrict file edits to specific directory |
/guard |
Full safety mode - destructive warnings + directory scoping |
/careful |
Safety guardrails for destructive commands |
/drawio |
Create diagrams, flowcharts, architecture diagrams |
/ink |
Ink terminal renderer for JSON specs |
/raycast |
Manage Raycast script commands |
/inference-profile |
AWS Bedrock inference profile details |
/find-skills |
Discover and install agent skills |
These are standalone skill files that serve as reference guides, not slash commands:
| File | Description |
|---|---|
code-review-senior-guidelines.md |
General senior-reviewer patterns for backend code |
codex-delegation.md |
How to delegate tasks to OpenAI Codex |
engineering-judgment.md |
"Spinach Rule" - when to push back on flawed assumptions |
generate-ralph-loop.md |
Auto-generate recurring task loop commands |
pr-workflow.md |
PR creation, review, and commit workflow template |
ui-testing.md |
UI testing reference (always use incognito mode) |
34 slash commands in commands/:
YAML-frontmatter markdown files that define prompts, tool permissions, and behaviors for specific workflows. Categories include PR review, deployment, research, monitoring, and productivity.
Lifecycle hooks with shell automation:
| Event | Hooks | Purpose |
|---|---|---|
| SessionStart | peon-ping, iterm2-claude | Sound notification, iTerm2 integration |
| Stop | peon-ping, iterm2-claude | Completion notification |
| Notification | peon-ping, iterm2-claude | Alert notification |
| UserPromptSubmit | hook-handle-use, iterm2-claude | Usage tracking |
- iterm2-claude.sh - iTerm2 tab color/badge integration based on session state
- peon-ping/scripts/hook-handle-use.sh - Sound pack switcher hook
Note: The main peon.sh script and sound files are not included (they reference local audio assets).
- general-purpose - Complex multi-step tasks
- Explore - Fast codebase exploration (quick/medium/very thorough)
- Plan - Software architect for implementation planning
- claude-code-guide - Claude Code / API documentation questions
- plugin-dev:skill-reviewer - Review skill quality
- plugin-dev:plugin-validator - Validate plugin structure
- plugin-dev:agent-creator - Create new agents
/orchestrate- Run multiple Claude/Codex agents in tmux panes/codex-swarm- Break tasks into subtasks across parallel Codex agents/swarm-review- 6 parallel review agents simultaneously- Agent teams via
CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1
{
"defaultMode": "auto",
"allow": [
"Bash(xargs cat:*)",
"Bash(git pull:*)",
"Bash(git show:*)"
]
}Custom Sanskrit mantra spinner verbs (108 mantras) replace default loading messages.
Custom bash script for status bar integration with iTerm2.
File-based persistent memory at ~/.claude/projects/<project>/memory/ with types:
- user - Role, goals, preferences
- feedback - Corrections and confirmed approaches
- project - Ongoing work context
- reference - Pointers to external systems
tstack/
README.md # This file
mcp.json.example # MCP server config template
settings.json.example # Claude Code settings template
skills/ # 45 skill directories + 6 reference docs
ship/SKILL.md
review/SKILL.md
investigate/SKILL.md
...
code-review-senior-guidelines.md
engineering-judgment.md
pr-workflow.md
...
commands/ # 34 slash command definitions
deploy-staging.md
review-pr.md
soc.md
...
hooks/ # Lifecycle shell scripts
iterm2-claude.sh
peon-ping/scripts/hook-handle-use.sh
mcp-servers/ # 7 custom MCP server implementations
shared/ # DB, kubectl, Teams webhook utilities
opensearch-direct/ # OpenSearch query server
opensearch-dashboards/ # OpenSearch dashboards proxy
prometheus/ # PromQL server
notion/ # Notion API server
soc-investigations/ # Example investigation queries
soc-infrastructure/ # Example pod health, OOMs, disk IO
soc-integrations/ # Example integration health monitoring
- Start with MCP servers - They give Claude real-world access. Start with
sequential-thinkingandpuppeteer, add domain-specific ones as needed. - Build skills incrementally - Each skill is just a markdown file with a prompt. Start with your most repeated workflows.
- Hooks are underrated -
UserPromptSubmithooks let you inject behavior on every prompt. Use them for auto-logging or safety checks. - Plugins extend the system - The PM skills suite alone adds 80+ product management capabilities.
- Memory persists context - Use it for user preferences and project context that shouldn't be re-discovered each session.
MIT