FCabiddu/pocket-it
Agent system for Claude Code: analyst, architect, task-writer, developer, reviewer, security and tester, with hooks and scripts that enforce the gates
field: unknown workflow: Planning & specsFrontend & designTesting & TDDGit & commitsPR / code review
stack: Shell
ARCHETYPE
Orchestrator
A bench of specialised subagents. The main agent mostly delegates.
Copy this rig
# review before running: this installs third-party code
$ npx degit FCabiddu/pocket-it/mvp-builder/.claude ./rig-pocket-it # inspect, then merge into .claude/ MCP servers are added to Claude Code at local scope; env vars are shown as YOUR_… placeholders — we never store values. Files are fetched with degit into a separate folder so you can review before merging.
$ npx degit FCabiddu/pocket-it/mvp-builder/.claude/skills/mvp-builder .claude/skills/mvp-builder $ curl -fsSL --create-dirs -o .claude/agents/analyst.md https://raw.githubusercontent.com/FCabiddu/pocket-it/main/flow/.claude/agents/analyst.md $ curl -fsSL --create-dirs -o .claude/agents/architect.md https://raw.githubusercontent.com/FCabiddu/pocket-it/main/flow/.claude/agents/architect.md $ curl -fsSL --create-dirs -o .claude/agents/developer.md https://raw.githubusercontent.com/FCabiddu/pocket-it/main/flow/.claude/agents/developer.md $ curl -fsSL --create-dirs -o .claude/agents/reviewer.md https://raw.githubusercontent.com/FCabiddu/pocket-it/main/flow/.claude/agents/reviewer.md $ curl -fsSL --create-dirs -o .claude/agents/security.md https://raw.githubusercontent.com/FCabiddu/pocket-it/main/flow/.claude/agents/security.md $ curl -fsSL --create-dirs -o .claude/agents/task-writer.md https://raw.githubusercontent.com/FCabiddu/pocket-it/main/flow/.claude/agents/task-writer.md $ curl -fsSL --create-dirs -o .claude/agents/tester.md https://raw.githubusercontent.com/FCabiddu/pocket-it/main/flow/.claude/agents/tester.md $ curl -fsSL --create-dirs -o .claude/agents/mvp-builder.md https://raw.githubusercontent.com/FCabiddu/pocket-it/main/mvp-builder/.claude/agents/mvp-builder.md
This rig commits no guardrails. Here is the community baseline instead — the deny/ask rules most often found across all 16.2k rigs:
{
"permissions": {
"deny": [
"Read(./.env)",
"Read(./.env.*)",
"Read(**/.env)",
"Read(~/.ssh/**)",
"Bash(git push --force:*)",
"Read(**/*.pem)",
"Read(.env)",
"Read(~/.aws/**)",
"Read(**/*.key)",
"Read(**/.env.*)",
"Bash(git push --force*)",
"Bash(rm -rf:*)",
"Bash(rm -rf *)",
"Bash(sudo *)",
"Bash(sudo:*)",
"Bash(rm -rf /)",
"Bash(git push -f:*)",
"Read(.env.*)",
"Bash(git reset --hard*)",
"Bash(rm -rf /*)"
],
"ask": [
"Bash(git push:*)",
"Bash(git push *)",
"Bash(gh pr merge *)",
"Bash(git commit:*)",
"Bash(rm *)",
"Bash(gh pr merge:*)",
"Bash(gh pr create *)",
"Bash(git reset --hard:*)",
"Bash(npm publish:*)",
"Bash(git rebase *)"
]
}
} Skills (1)
Subagents (8)
| analyst model: sonnet | Turns a request ("I need this") into a real requirement, asking the user whatever is unclear. Knows the code well enough to say what is feasible. Works alongside the architect; writes no code and no t |
| architect model: opus | Technical owner of the whole project. Knows all of it, makes every technical choice (patterns, technologies, approaches, methods), keeps them in one directives file every developer follows, and design |
| developer model: sonnet | Develops one atomic task of any kind (frontend, backend, test) following the task and the architect's directives, runs the tests of what it touched, opens a PR, and fixes the reviewer's feedback. |
| reviewer model: sonnet | Checks a PR against its task (functional) and the architect's directives (technical), sends numbered feedback to fix, re-verifies after the fix. Never merges. |
| security model: opus | Looks at a project the way an attacker would, against the security principles of the flow. Audits a whole project once at its start, and checks every wave before it reaches main. Writes no app code; r |
| task-writer model: sonnet | Turns the requirement and the technical design of a feature into atomic, well-structured tasks — clear work, acceptance criteria, requirements and tests — grouped into waves of tasks that can run in p |
| tester model: sonnet | At the end of a wave, on the wave branch with all its PRs merged, writes the E2E tests the wave's tester task asks for, runs them once together with the existing E2E that cover the files the wave chan |
| mvp-builder model: sonnet | Full-stack MVP builder. Takes a plain-English description and delivers an award-winning static site — no frameworks, no build step, just an index.html plus a css/ folder (one file per section) and min |
Similar rigs
bpstack/harness
A reusable foundation for AI projects: rules, agents, templates, and commands.
Pragmatist 3.5k tok · · Q5
joshido/ai-template
A project-agnostic four-agent AI development workflow template for Claude Code, Gemini CLI, and GitHub Copilot.
Pragmatist 1.3k tok · · Q20
vannt-dev/aiws-factory
AIWS – AI Software Factory: turns a requirement into tested code through gated phases (analysis → design + test spec → human approval → per-task implementation with unit tests → review → PR). Language-agnostic, Claude Code first.
Fort Knox 1.5k tok · · Q19
vannt-dev/governed-agent-sdlc
Ship with agents. Keep humans in control. Cross-platform governance toolkit for AI coding agents.
Fort Knox 632 tok · · Q24