Skip to content
Back to skills

Dogfood Ralph

ASecurity

Dogfood ralph on this repo: run a ralph loop to build a ralph spec or feature, with this repo's setup and test gotchas. Use when asked to dogfood or to have ralph implement ralph work.

  • 3,203 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added November 12, 2026
ai-agentsgogitbackend

Works with

  • cli

Security analysis

A100/100

Scanned November 12, 2026

npx -y skills add mikeyobrien/ralph-orchestrator --skill dogfood-ralph --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Dogfood Ralph?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Dogfood Ralph
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/mikeyobrien-dogfood-ralph/badge)](https://www.skillsdirectory.com/skills/mikeyobrien-dogfood-ralph)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: dogfood-ralph
description: "Dogfood ralph on this repo: run a ralph loop to build a ralph spec or feature, with this repo's setup and test gotchas. Use when asked to dogfood or to have ralph implement ralph work."
metadata:
  internal: true
---

# Dogfood ralph on this repo

Launching, monitoring, measuring, and stepping in follow `skills/ralph-loop/references/headless-runs.md`. This skill adds only what is specific to running ralph inside its own repository.

## Setup

1. Cut a worktree for the work: `git worktree add ../ralph-<name> -b <branch> origin/main`. Commit the spec the loop will implement into that branch. This repo's `CLAUDE.md` says to implement only from an approved spec, so the loop needs it on disk.
2. Build the ralph that will run the loop, and copy the binary out of `target/`, so the loop's own `cargo build` can't swap it mid-run: `cargo build --release -p ralph-cli && cp target/release/ralph <outside>/ralph`.
3. Put the core config (`cli.backend`, `event_loop` limits) and `PROMPT.md` outside the worktree, and pass them with `-c` and `-P`.

**Done when** `<outside>/ralph doctor -c <outside>/run.yml -H builtin:<preset>` passes in the worktree and the spec commit is on the branch.

## Write these into the loop's prompt

The loop's hats read the prompt, not this skill:
- The verification commands: `cargo test -p ralph-core`, `cargo test -p ralph-adapters`, `cargo test -p ralph-cli -- --test-threads=1`, `cargo clippy --workspace --all-targets`, `cargo fmt --all --check`, and `./scripts/sync-embedded-files.sh check` after preset or `crates/ralph-core/data/` edits.
- That ralph-cli's wave and backend tests race on `PATH` when run in parallel, which also happens on `main`. Without this, the Critic rejects for failures that aren't regressions.

## While it runs

Tests that call `ralph emit` without clearing `RALPH_EVENTS_FILE` write stray topics (`test.topic`, `t.x`, `t.y`) into the live loop's events file whenever a hat runs the suite. They route nowhere. Filter them out of your monitor.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

View all in ai-agents →

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.