# skill-pipeline

skill-pipeline sizes up a coding task, then decides which skills to switch on: nothing extra for a typo, planning plus checks and failure recovery for a big refactor.

- skill · `type=skill` · Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, OpenCode, Windsurf (Devin Desktop), Claude.ai · 114 installs · ★ 341 repo stars
- Author: pskoett
- License: not stated · policy=link-only · description only; the file stays with the author
- Source: https://github.com/pskoett/pskoett-ai-skills/tree/HEAD/skills/skill-pipeline/SKILL.md
- Updated: 2026-10-19
- Security check: automated, 2026-11-16 (text) — no signs of malicious code found.
- HTML: https://affpapa.org/en/nevibecode/skills/skill-pipeline

## What it does

Think of it as the conductor. It writes nothing into the project itself; its output is routing. Every task gets a size first: typo, single-file fix, feature across a few files, big refactor, multi-session work, batch of tasks. A typo needs no skills at all. A small fix gets `verify-gate` and `simplify-and-harden`. Medium adds `intent-framed-agent` and `session-context`. Large runs the full sequence and suggests opening with `/plan-interview`.

One loop lives inside a session: frame the intent, pull bounded context, implement, check compile and tests, recover through `self-healing` when a command or build fails. The other loop spans sessions: `learning-aggregator` looks for repeats in `.learnings/`, `self-improvement` turns the human-approved ones into project instructions, `eval-creator` writes them as standing tests. The `pre-flight-check` hook brings earlier findings back at session start. Batch work is cut into bounded units with a gate after each. In CI, with `CI=true` or `GITHUB_ACTIONS=true` set, the pipeline switches to a headless pass over the files a PR touches.

The companion skills have to be there first. `skill-pipeline` dispatches to `verify-gate`, `self-healing` and the rest rather than replacing them, so routing with none of them installed goes nowhere. Edge cases sit in `references/classification-rules.md`.

## Install

### Claude Code (native platform)

1. **Skills CLI (Vercel)**
```bash
npx skills add pskoett/pskoett-ai-skills --skill skill-pipeline -a claude-code
```

2. **Manually: put the files in place**
- File: `~/.claude/skills/skill-pipeline/SKILL.md` — Copy the whole skill folder; for one project use .claude/skills/skill-pipeline/

### Codex

1. **Skills CLI (Vercel)**
```bash
npx skills add pskoett/pskoett-ai-skills --skill skill-pipeline -a codex
```

2. **Manually: put the files in place**
- File: `.agents/skills/skill-pipeline/SKILL.md` — For all projects use ~/.agents/skills/

### Cursor

1. **Skills CLI (Vercel)**
```bash
npx skills add pskoett/pskoett-ai-skills --skill skill-pipeline -a cursor
```

2. **Manually: put the files in place**
- File: `.cursor/skills/skill-pipeline/SKILL.md` — Cursor also reads .agents/skills/ and .claude/skills/

### GitHub Copilot

1. **GitHub CLI**
```bash
gh skill install pskoett/pskoett-ai-skills skills/skill-pipeline
```

2. **Manually: put the files in place**
- File: `.github/skills/skill-pipeline/SKILL.md`

### Gemini CLI

1. **Skills CLI (Vercel)**
```bash
npx skills add pskoett/pskoett-ai-skills --skill skill-pipeline -a gemini-cli
```

2. **Platform CLI**
```bash
gemini skills install https://github.com/pskoett/pskoett-ai-skills --path skills/skill-pipeline
```

### Windsurf (Devin Desktop)

**Manually: put the files in place**
- File: `.devin/skills/skill-pipeline/SKILL.md` — The legacy .windsurf/skills/ path still works

### OpenCode

**Skills CLI (Vercel)**
```bash
npx skills add pskoett/pskoett-ai-skills --skill skill-pipeline -a opencode
```

### Claude.ai

**Claude.ai**
- Zip the folder and upload it in Settings → Capabilities → Skills

Where a file is given instead of a command, copy the files from the source to that path: https://github.com/pskoett/pskoett-ai-skills/tree/HEAD/skills/skill-pipeline/SKILL.md

## When to use it
- split a large feature into bounded units with a gate after each
- match the depth of checking to the size of the task
- carry past fixes and findings into a fresh session
- recover when a test or build breaks mid-task
- turn recurring mistakes into a project rule

## Example prompts
- Classify this task and pull in the skills it needs — one file changed, but the logic is risky.
- This will span several sessions: cut it into bounded units and gate each one.
- Run the outer loop: pull patterns out of `.learnings/` and propose what to promote.

## Similar
- [tdd-migration-pipeline](https://affpapa.org/en/nevibecode/skills/tdd-migration-pipeline.md) — skill · 480 installs
- [do](https://affpapa.org/en/nevibecode/skills/do.md) — skill · 7500 installs
- [figure-it-out](https://affpapa.org/en/nevibecode/skills/figure-it-out.md) — skill · 10300 installs
- [verify](https://affpapa.org/en/nevibecode/skills/yeachan-heo-verify.md) — skill · 721 installs
- [plan](https://affpapa.org/en/nevibecode/skills/hyperb1iss-plan.md) — skill · 606 installs
- [loop-architect](https://affpapa.org/en/nevibecode/skills/loop-architect.md) — skill · ★ 141 repo stars

From the same set: [pskoett/pskoett-ai-skills](https://affpapa.org/en/nevibecode/by/pskoett/pskoett-ai-skills.md) — 10 cards

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md
