danielfarnose/claude-code-dev-loop
A multi-agent development loop for Claude Code: five specialized subagents plan, implement, verify and merge — with a QA gate that cannot edit the code it judges.
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 danielfarnose/claude-code-dev-loop/agents ./rig-claude-code-dev-loop/agents $ npx degit danielfarnose/claude-code-dev-loop/commands ./rig-claude-code-dev-loop/commands $ npx degit danielfarnose/claude-code-dev-loop/skills ./rig-claude-code-dev-loop/skills
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 danielfarnose/claude-code-dev-loop/skills/codex-board .claude/skills/codex-board $ npx degit danielfarnose/claude-code-dev-loop/skills/codex-explain .claude/skills/codex-explain $ npx degit danielfarnose/claude-code-dev-loop/skills/codex-patrol .claude/skills/codex-patrol $ npx degit danielfarnose/claude-code-dev-loop/skills/codex-run .claude/skills/codex-run $ npx degit danielfarnose/claude-code-dev-loop/skills/codex-security-audit .claude/skills/codex-security-audit $ npx degit danielfarnose/claude-code-dev-loop/skills/inspect-project .claude/skills/inspect-project $ npx degit danielfarnose/claude-code-dev-loop/skills/qa-live-user .claude/skills/qa-live-user $ npx degit danielfarnose/claude-code-dev-loop/skills/recording-learnings .claude/skills/recording-learnings
$ curl -fsSL --create-dirs -o .claude/agents/architect.md https://raw.githubusercontent.com/danielfarnose/claude-code-dev-loop/main/agents/architect.md $ curl -fsSL --create-dirs -o .claude/agents/developer.md https://raw.githubusercontent.com/danielfarnose/claude-code-dev-loop/main/agents/developer.md $ curl -fsSL --create-dirs -o .claude/agents/legal.md https://raw.githubusercontent.com/danielfarnose/claude-code-dev-loop/main/agents/legal.md $ curl -fsSL --create-dirs -o .claude/agents/pm.md https://raw.githubusercontent.com/danielfarnose/claude-code-dev-loop/main/agents/pm.md $ curl -fsSL --create-dirs -o .claude/agents/qa.md https://raw.githubusercontent.com/danielfarnose/claude-code-dev-loop/main/agents/qa.md $ curl -fsSL --create-dirs -o .claude/agents/security.md https://raw.githubusercontent.com/danielfarnose/claude-code-dev-loop/main/agents/security.md
This rig commits no guardrails. Here is the community baseline instead — the deny/ask rules most often found across all 7,204 rigs:
{
"permissions": {
"deny": [
"Read(./.env)",
"Read(**/.env)",
"Read(~/.ssh/**)",
"Bash(rm -rf *)",
"Read(**/*.pem)",
"Bash(rm -rf /)",
"Bash(git push --force:*)",
"Bash(sudo *)",
"Read(.env)",
"Bash(rm -rf /*)",
"Read(./.env.*)",
"Read(~/.aws/**)",
"Bash(git push --force*)",
"Bash(rm -rf:*)",
"Read(**/*.key)",
"Read(**/.env.*)",
"Bash(sudo:*)",
"Bash(git reset --hard*)",
"Bash(git reset --hard:*)",
"Read(.env.*)"
],
"ask": [
"Bash(git push:*)",
"Bash(git push *)",
"Bash(git commit:*)",
"Bash(rm *)",
"Bash(rm:*)",
"Bash(git rebase *)",
"Bash(wget *)",
"Bash(npm publish:*)",
"Bash(git commit *)",
"Bash(gh pr merge *)"
]
}
} Skills (8)
codex-boardcodex-explaincodex-patrolcodex-runcodex-security-auditinspect-projectqa-live-userrecording-learnings
Subagents (6)
| architect model: claude-fable-5 | Analyzes a CODE task in the current project, does brainstorming + a minimalist plan (ponytail), and generates a ticket in the project's tickets path (defined in .claude/squad.md). Does NOT write featu |
| developer model: claude-sonnet-5 | Implements the ticket with TDD and minimum code (ponytail) on the current project's real stack (defined in .claude/squad.md). Receives a ticket path and produces verified, committed code. Use it after |
| legal model: claude-opus-5 | Legal-texts writer for a project going to production. Renders the plugin's generic templates (templates/legal/*, no product inside) with what the repo already says plus the operator's answers to what |
| pm model: claude-opus-5 | Product manager of the current project's development squad. Keeps the backlog prioritized, turns the human's raw ideas into product specs that feed the architect in two modes (discovery: intent + open |
| qa model: claude-sonnet-5 | Gate of the development loop. Read-only over the code. Runs the current project's verification (defined in .claude/squad.md) and answers APPROVED / REJECTED with actionable reasons. Use it after the d |
| security model: claude-opus-5 | MANUAL security auditor (outside the dev loop). Scans the current project's code for vulnerabilities, focused on multi-tenant isolation (one client must not see another's data), data/pricing leaks, pr |
Slash commands (5)
/board/explain/launch/patrol/run
Similar rigs
secretj/secretj-claude-config
개인 Claude Code dotfiles — agents, commands, skills, hooks, memory, MCP 템플릿. clone + install.sh 로 환경 복원.
Orchestrator 1.8k tok ·
yeyo11/agentry
Run Claude Code as a service: REST API, web UI and multi-agent orchestration in one container
Pragmatist 5.2k tok ·
IntuitivePhella/AI-PhellOS
Portable AI development workflow template and CLI for spec-driven, test-aware software delivery with Codex, Claude Code, and other coding agents.
Orchestrator 3.1k tok ·
ThinhHoang1/claude-code-agent-team
An AI dev team for Claude Code agent teams: 8 role agents, a HANDOFF protocol and 3 hook quality gates — with the TeammateIdle infinite-loop bug fixed and regression-tested.
Orchestrator 1.6k tok ·