stop prompting; start shipping¶
Professional development workflows for AI coding assistants. Skip the iteration loops. Keep work attached to projects, artifacts, and feedback instead of ephemeral sessions. 40 workflow skills. 51 specialized agents. Use the same workflows in Claude Code, OpenCode, and Codex, then review them in Arcade.
-
Project overview
-
Workflow detail
-
Artifact feedback
-
Claude Code workflow
Start with an outcome¶
Pick the workflow that matches the outcome you want. RP1 keeps the artifact trail and review loop intact across Claude Code, OpenCode, and Codex.
Ship a feature
Turn a feature idea into requirements, design, implementation, and verification with durable artifacts at every step.
/build "user-auth"Make a quick change
Push through a bounded change fast without giving up structure, reviewability, or resumability.
/build-fast "Add dark mode toggle"Power through small changes
Work through a queue of small requests with speed, but still grounded in the codebase, its patterns, and the knowledge base.
/speedrun "Tighten the empty state copy"Review a PR
Get structured findings, sharper confidence, and visual summaries instead of a shallow skim.
/pr-review "feature/auth"Investigate a bug
Move from a vague failure to an evidence-backed root cause instead of bouncing through ad hoc prompts.
/code-investigate "bug-123" "Login fails"Build project context
Generate a project-aware knowledge base so every workflow starts from your actual system, not generic assumptions.
/knowledge-buildDeep research
Investigate a system, compare multiple projects, or pull together a technical brief with grounded findings.
/deep-research "Compare backend and frontend error handling"Same workflow, different harness
/build "user-auth"
/rp1-dev-build "user-auth"
$rp1-dev-build "user-auth"
Why rp1?¶
Ship features without iteration loops
Constitutional prompts encode expert patterns with built-in rules. No "let me revise that" — tasks complete in one shot.
Project-aware from the first command
Run knowledge-build once. Your architecture becomes context for every command. No generic advice — everything respects your patterns.
Give feedback in context
Open artifacts in Arcade and annotate the exact file or section you want changed. Agents can pick that feedback up without forcing you into copy-paste loops.
Catch bad assumptions before coding
Automatic hypothesis testing catches bad assumptions early. Design decisions get validated against your codebase before implementation begins.
Switch harnesses without switching costs
rp1 keeps the workflow layer stable across Claude Code, OpenCode, and Codex. The syntax adapts to each host, but the workflow, artifacts, and review loop stay the same.
Durable artifacts, not disposable chats
Requirements, design, tasks, verification, and reports stay attached to the project and remain resumable across sessions, reviews, and handoffs.
rp1 Principles¶
Batteries Included
Workflows, subagents, KB generation, and structured artifacts ship together. No prompt assembly required.
Always Open Source
Fully pluggable into existing agentic tools. Your workflows, your control.
Observable by Default
Runs, gates, artifacts, and annotations stay visible in Arcade instead of disappearing into chat scrollback.
Stable Workflows, Evolving Harnesses
Keep the workflow layer stable while frontier models and host tools keep changing.
Same Workflows, Different Hosts
Use rp1 from Claude Code, OpenCode, or Codex without rewriting how your team works.
Guides¶
PR Review
Thorough multi-pass analysis that catches what single-reviewer approaches miss. Visual diagrams show exactly what changed.
Team Onboarding
Get new developers productive on your codebase fast. Knowledge base + guided exploration beats reading docs alone.