agentleFS
shubham0704shubham0704/claude-skills/rpi-workflow/SKILL.md

rpi-workflow

Skill2 starsChanged 6 months agoMITView on GitHub

Research-Plan-Implement workflow with Codex peer review. Use when the user says /rpi, asks to start a research-plan-implement cycle, wants Codex review on a plan, or needs to create research docs, plans, or implementation reports following the RPI protocol.

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 5 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    All artifacts go in docs/ at the project root:

    line 20
  2. Distinctive:

    Run from the project root:

    line 59
  3. Very specific:

    Always use --skip-git-repo-check for non-git repos

    line 66
  4. Very specific:

    Read docs/plans/FEEDBACK-<repo>-<task>.md and present key findings to the user:

    line 72
  5. Distinctive:

    Research first: Never plan without understanding the codebase

    line 96
  6. Distinctive:

    Parallel agents: Use multiple Explore agents simultaneously

    line 97
  7. Very specific:

    Compact findings: Use file:line references, not full code dumps

    line 98
  8. Distinctive:

    Questions for Codex: Always include 3-5 specific architectural questions

    line 99
  9. Distinctive:

    Parity checks: Before deleting old code, run side-by-side comparisons

    line 100

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

RPI Workflow: Research -> Plan -> Implement

A structured workflow for complex engineering tasks with Codex architectural review.

When to Invoke

  • User says /rpi or "start RPI"
  • User wants structured research before implementation
  • User wants Codex to review a plan
  • User needs to create research docs or implementation plans

File Structure

All artifacts go in docs/ at the project root:

docs/
├── research/
│   └── YYYY-MM-DD-topic.md        # Research findings
└── plans/
    ├── PLAN-<repo>-<task>.md      # Claude creates
    ├── FEEDBACK-<repo>-<task>.md  # Codex creates
    ├── IMPL-<repo>-<task>.md      # Claude creates after implementation
    └── VALIDATION-<repo>-<task>.md # Codex creates (optional)

Workflow Steps

Step 1: Research Phase

Launch parallel Explore agents to gather codebase understanding:

Task(subagent_type="Explore", prompt="Research how X works...")
Task(subagent_type="Explore", prompt="Research how Y connects to Z...")

Save findings to docs/research/YYYY-MM-DD-topic.md using the research template. For the template structure, see TEMPLATES.md.

Step 2: Plan Phase

Create docs/plans/PLAN-<repo>-<task>.md with:

  • Goal (1-2 sentences)
  • Key Files (path:line references)
  • Problem (what's broken/missing)
  • Proposed Solution (phased, with file-level changes)
  • Success Criteria (checkboxes)
  • Risks (numbered)
  • Questions for Review (for Codex)

Step 3: Codex Review

Run from the project root:

codex exec --skip-git-repo-check "Read docs/plans/PLAN-<repo>-<task>.md and provide architectural feedback. Review the proposed solution, identify risks, suggest improvements. Write your feedback to docs/plans/FEEDBACK-<repo>-<task>.md"

Notes:

  • Always use --skip-git-repo-check for non-git repos
  • No --quiet flag exists
  • Codex writes feedback autonomously to the FEEDBACK file

Step 4: Read & Incorporate Feedback

Read docs/plans/FEEDBACK-<repo>-<task>.md and present key findings to the user:

  • Executive assessment
  • Missing edge cases
  • Better abstractions suggested
  • Risks not identified
  • Phase reordering suggestions
  • Actionable recommendations

Step 5: Implement

Make the changes. Create docs/plans/IMPL-<repo>-<task>.md documenting:

  • What was actually changed (with path:line refs)
  • Deviations from plan and why
  • Test results
  • Remaining work

Step 6: Validate (Optional)

codex exec --skip-git-repo-check "Read docs/plans/IMPL-<repo>-<task>.md and validate the implementation. Check if success criteria are met, identify any issues. Write your validation to docs/plans/VALIDATION-<repo>-<task>.md"

Guidelines

  • Research first: Never plan without understanding the codebase
  • Parallel agents: Use multiple Explore agents simultaneously
  • Compact findings: Use file:line references, not full code dumps
  • Questions for Codex: Always include 3-5 specific architectural questions
  • Parity checks: Before deleting old code, run side-by-side comparisons
  • Phase boundaries: Each phase should be independently mergeable

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.