From dev-loop
Scaffold a Rails project to use /dev-loop. Verifies @kurenn/roundhouse and @openai-codex/codex are installed, adds a '## Dev-loop config' block to CLAUDE.md (gate, critical paths, extra rating axes, learnings path, PR conventions), and creates the learnings file. Run once per repo before the first /dev-loop.
How this skill is triggered — by the user, by Claude, or both
Slash command
/dev-loop:dev-loop-setupThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
One-time setup. Make the smallest set of changes that lets `/dev-loop` run cleanly in
One-time setup. Make the smallest set of changes that lets /dev-loop run cleanly in
this repository, then tell the user what (if anything) they still need to do.
Report whether each is installed; don't fail if missing (the loop degrades), but tell the user what they lose:
ls -d ~/.claude/plugins/cache/kurenn/roundhouse/*/ 2>/dev/null # /rails-feature, /rails-bugfix (Phase 2 army)
ls -d ~/.claude/plugins/cache/openai-codex/codex/*/ 2>/dev/null # /codex:adversarial-review (Phase 3)
/plugin install roundhouse@kurenn./plugin install codex@openai-codex.## Dev-loop config block in CLAUDE.mdIf the repo has no CLAUDE.md, create one. If it has one but no ## Dev-loop config
section, append the block below. If the section already exists, leave it and show the
user the current values. Fill the placeholders by inspecting the repo — don't leave
<...> in the file:
none.design-system fidelity if the repo has a component
library / design system; otherwise none.gh pr edit; assets-branch pattern).## Dev-loop config
Knobs read by the `/dev-loop` skill (@kurenn/dev-loop). Defaults shown; edit to taste.
- **Gate:** overall ≥ 8.5 AND every axis ≥ 7
- **Base rating axes:** correctness, simplicity, test coverage, naming, performance risk, security risk
- **Extra rating axes:** <e.g. design-system fidelity — or `none`>
- **Critical paths:** <e.g. money / auth / KYC — these axes also require ≥ 8.5 — or `none`>
- **Learnings file:** docs/dev-loop-learnings.md
- **Fix-round cap:** 2
- **PR conventions:** <repo-specific screenshot/embed/gh quirks — or `standard gh pr create`>
If the configured learnings file (default docs/dev-loop-learnings.md) doesn't exist,
create it with this header so Phase 6 has somewhere to append:
# Dev-loop learnings
Durable, reusable insight captured by `/dev-loop` (Phase 6). Append here — newest first.
Map, not diary. Nothing user-specific or secret.
## Entry format
### YYYY-MM-DD — <short title>
**Context:** <feature/bug + PR # if any>
**Learning:** <the durable, reusable insight>
**How to apply:** <what to do next time>
Tell the user: what was created/edited, which prerequisites are missing (with the exact
/plugin install commands), and that they can now run /dev-loop <task>.
npx claudepluginhub kurenn/dev-loopProvides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.