Refine
The front door for any new ask — feature, bug, spike, or chore. Use when the product owner has a raw idea in their head (or a seed in Roadmap/00-ideas/seeds) and wants to turn it into shippable, sliced work. Runs orientation → appetite → classification → "can we already do this?" → disambiguation → platform-first reframe → slicing, lands a Definition- of-Ready pitch in 00-ideas/seeds, and on approval funds the bet, scaffolds + commits the epic + sprint docs and emits the builder kickoff — epic-mode by default (one orchestrated run across the whole epic), per-sprint only as the named exception.
Install
npx skills add danybgoode/golden-frijoles --skill refinegit clone --depth 1 https://github.com/danybgoode/golden-frijoles.git /tmp/golden-frijoles
mkdir -p ~/.claude/skills
cp -r /tmp/golden-frijoles/skills/plugins/golden-frijoles/skills/refine ~/.claude/skills/refinePersonal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.
From the SKILL.md
# Refine — the planning front door (Cowork)
> **Running this skill's scripts.** Each `node scripts/<name>.mjs <args>` below means, from the project root:
> `if [ -f scripts/<name>.mjs ]; then node scripts/<name>.mjs <args>; else npx -y @golden-frijoles/kit@1.17.5 <name> <args>; fi`
> The project's own copy always wins. The kit needs no install and copies nothing into the repo. If npx fails on
> the network or the registry (ENOTFOUND, ECONNREFUSED, ETIMEDOUT, E404, a proxy error), report **could not
> look: the kit is unreachable**, not a broken project, and offer to retry online or to copy the script into
> `scripts/`. This covers the scripts the kit carries (`frijoles-kit --list`). Any other `scripts/` command here is
> project tooling: if the project lacks it, say so and stop. Never re-implement a script's logic inline.
> If a script prints `GF-NEEDS-SETTING {"key",…}`, ask the user that question once, save the answer with
> `npx -y @golden-frijoles/kit@1.17.5 config set <key> <value>` (always the kit, never local-wins: a project
> may own an unrelated `scripts/config.mjs`), tell them they can change it later the same way, and run the
> script again if it exited 7.
> **Role split this skill assumes:** *Cowork plans, Claude Code builds.* This skill is the
> Cowork half. It produces and edits **product docs under `Roadmap/`** and **nothing else** —
> no code, no `tasks/` engineering log (that's Claude Code's lane). The handoff is file-based:Details
- Skill name
refine- Repository
- danybgoode/golden-frijoles
- Path
skills/plugins/golden-frijoles/skills/refine/SKILL.md- Repo updated
- 17 November 2026
- Listed in
- karanb192/awesome-claude-skills
FAQ
What does the Refine skill do?
The front door for any new ask — feature, bug, spike, or chore. Use when the product owner has a raw idea in their head (or a seed in Roadmap/00-ideas/seeds) and wants to turn it into shippable, sliced work. Runs orientation → appetite → classification → "can we already do this?" → disambiguation → platform-first reframe → slicing, lands a Definition- of-Ready pitch in 00-ideas/seeds, and on approval funds the bet, scaffolds + commits the epic + sprint docs and emits the builder kickoff — epic-mode by default (one orchestrated run across the whole epic), per-sprint only as the named exception.
How do I install the Refine skill?
skills CLI (Claude Code, Cursor, Codex and more): npx skills add danybgoode/golden-frijoles --skill refine. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.
Related skills
Babysit PR
danybgoode/golden-frijoles
Advisory PR watch for a single open PR — retries flaky CI (re-runs failed workflow runs) and surfaces merge conflicts via one comment; never merges, never a required check. Use when the product owner asks to "babysit this PR", "check on open PRs", "retry flaky CI", "is this PR stuck", "why isn't this PR moving", or as the nightly ops routine's third step (once per open PR across the project's repos). Runs scripts/babysit-pr.mjs, which does the gh reads, the retry, and the comment post. A clean PR (no conflict, no failing checks) gets NO comment — this tool never adds nightly noise to a healthy
Build Order Sync
danybgoode/golden-frijoles
Regenerates Roadmap/00-ideas/BUILD-ORDER.md when it has drifted from the SSOT (each epic README's frontmatter status:) and opens a claude/ docs PR with the fix — never hand-edits the board. Use when the product owner asks to "sync the build order", "regenerate BUILD-ORDER.md", "check board drift", "is the build-order board stale", "open a build-order PR", or as the nightly ops routine's first step. Runs scripts/build-order-sync.mjs, which does the check, the regen, and (on real drift) the branch/commit/ push/PR — this skill just invokes it and reports the result.
Doc Hygiene
danybgoode/golden-frijoles
Keeps the always-read session-start docs (this project's AGENTS.md, Roadmap/WAYS-OF-WORKING.md, Roadmap/LEARNINGS.md, Roadmap/README.md poster) from re-bloating after the one-time de-noise sweep. Use when the product owner asks to "check doc hygiene", "measure LEARNINGS size", "is LEARNINGS bloated again", "run the doc hygiene report", or as the weekly Routine C hygiene pass. Runs scripts/doc-hygiene.mjs, reviews its flagged candidates against the source docs, and emits a dated advisory report. Never edits Roadmap docs — proposals only, the product owner or the builder hand-merges.
Tracking Service Reliability
jeremylongshore/claude-code-plugins-plus-skills
This skill enables Claude to define and track Service Level Agreements (SLAs), Service Level Indicators (SLIs), and Service Level Objectives (SLOs) for improved service reliability. It is triggered when the user needs to establish, monitor, or analyze service performance metrics. Use this skill when the user mentions "SLA", "SLI", "SLO", "error budget", "service reliability", or "track service performance". The skill helps to define key metrics, set targets, and monitor performance against those targets.
Building Neural Networks
jeremylongshore/claude-code-plugins-plus-skills
This skill allows Claude to construct and configure neural network architectures using the neural-network-builder plugin. It should be used when the user requests the creation of a new neural network, modification of an existing one, or assistance with defining the layers, parameters, and training process. The skill is triggered by requests involving terms like "build a neural network," "define network architecture," "configure layers," or specific mentions of neural network types (e.g., "CNN," "RNN," "transformer").
Constant Time Analysis
5lucaspmarie-a11y/claude-skills-vault
Analyze cryptographic code to detect operations that leak secret data through execution timing variations.
Browse the most popular Claude skills, all finance & crypto skills, or write your own with the skills generator.