Stats
Actions
Tags
From buidl
Suggests appropriate Loop commands (/buidl, /buidl-spec, /buidl-review, /buidl-resume) to turn feature ideas into PRs via challenge, spec, build, and review cycles. Useful for new projects, planning, or resuming workflows.
How this skill is triggered — by the user, by Claude, or both
Slash command
/buidl:loop-guideThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
The Loop is a development lifecycle plugin. It turns ideas into production-ready PRs through structured interrogation, specification, and autonomous build-review cycles. v3 adds checkpointing, cost tracking, resume capability, and a learning system.
The Loop is a development lifecycle plugin. It turns ideas into production-ready PRs through structured interrogation, specification, and autonomous build-review cycles. v3 adds checkpointing, cost tracking, resume capability, and a learning system.
/buidl "idea description"Suggest when the user:
This runs the FULL pipeline: challenge the idea -> generate a spec -> explore the codebase -> build in isolation -> review -> iterate until clean -> retrospective -> produce a PR.
/buidl path/to/spec/Suggest when the user:
/buidl-spec and now wants to buildThis skips the challenge and specify phases and goes straight to explore -> build -> review.
/buidl-spec "idea description"Suggest when the user:
This runs ONLY the challenge and specify phases, producing spec documents without building anything.
/buidl-review <PR-number>Suggest when the user:
/buidl-resumeSuggest when the user:
/buidl-resumeReads the checkpoint and state file to continue from the last phase.
/buidl-statusSuggest when the user asks about the current loop state or seems unsure where things stand. Shows tokens used, elapsed time, and checkpoint info.
/buidl-cancelSuggest when the user wants to stop a running loop but keep the work.
/buidl-cleanSuggest when the user wants to stop AND remove the worktree and branch.
Challenge -- The Loop interrogates your idea. It runs through 5 gates (goal alignment, build vs buy, simplest thing, problem framing, testability), asks structured spec questions, then runs a pre-mortem and plays devil's advocate. This catches gaps before they become bugs. Checkpoint saved.
Specify -- From your answers, it generates three documents: requirements.md (what), design.md (how), tasks.md (work breakdown). It validates these against a quality checklist and asks you to approve before building. Checkpoint saved.
Explore -- Two agents analyze the codebase in parallel: one maps the structure and conventions, one finds code relevant to the spec. This gives the builder context. Checkpoint saved. Cost logged.
Build -- Specialized agents work in an isolated git worktree (your main branch is never touched). For OPNet: contract-dev, frontend-dev, backend-dev, auditor, deployer, and UI tester each handle their domain. For generic projects: dynamic domain agents are generated from templates. Each agent has a max_turns limit. Checkpoint after each agent. Cost logged per agent.
Review -- A reviewer agent reads the PR diff (read-only, can't touch code). It checks spec compliance, correctness, security, test quality, and conventions. If it fails, findings are routed back to responsible agents for the next cycle. Checkpoint saved.
Wrap-up -- A retrospective is generated and saved to the learning store. Future sessions consult past retrospectives for matching project types.
--max-cycles N (default 3) -- how many build-review cycles before stopping--max-retries N (default 5) -- how many verify-fix attempts per cycle--max-tokens N (optional) -- token budget; if exceeded, asks to continue or cancel--skip-challenge -- skip the interrogation phase (use when you know exactly what you want)--builder-model opus|sonnet -- which model runs the builder--reviewer-model opus|sonnet -- which model runs the reviewernpx claudepluginhub bc1plainview/buidl-opnet-pluginGenerates feature specs, implementation plans with task checklists, and project loop infrastructure via interactive user interviews on scope, risk tolerance, and validation.
Orchestrates autonomous end-to-end project building from a description via CodeClaw pipeline: ideas, tasks, releases, implementation, docs, and social announcements.
Guides developers through the full development pipeline: discover, brainstorm, plan, execute, review, and ship. Invoke when starting work on a bug, feature, improvement, or task.