Onboard the harness into an existing (legacy / non-harness) codebase — survey it, document it, wire init.sh to its own build/test, init harness without clobbering, and seed a backlog.
Scanned 9/8/2026
Install to Claude Code
npx -y skills add nobodyonlyc/skills --skill workflow-onboard --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Workflow Onboard?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/nobodyonlyc-workflow-onboard)More formats (shields.io, HTML) on the badges page.
---
name: workflow-onboard
description: Onboard the harness into an existing (legacy / non-harness) codebase — survey it, document it, wire init.sh to its own build/test, init harness without clobbering, and seed a backlog.
---
Onboard target: $ARGUMENTS
Multi-agent workflow that brings an **existing codebase under harness control** without rewriting it. This is the bridge between greenfield ([workflow-bootstrap](../workflow-bootstrap/SKILL.md)) and adding work to an already-harnessed repo ([Route 3](../workflow-intake/references/route-3-add-feature.md)): the project is real and active, but has never had the harness. Each phase has a playbook in [`references/`](references/); see [`examples/`](examples/) for a full worked run.
## Skills this workflow drives
- [core-explain](../core-explain/SKILL.md) — read and map the existing code, patterns, and build/test commands (Phase 1).
- [plan-us-backlog-generator](../plan-us-backlog-generator/SKILL.md) — seed the backlog from current state + the requested work (Phase 3).
- [dev-db-designer](../dev-db-designer/SKILL.md) — consulted only if the survey finds a database whose schema must be documented.
## Phases
1. **Survey & discovery** → [references/phase-1-survey.md](references/phase-1-survey.md)
Fan-out readers map the repo: languages, frameworks, entry points, and the project's **own** build/test/run commands. Output a survey report.
2. **Documentation & baseline** → [references/phase-2-docs-and-baseline.md](references/phase-2-docs-and-baseline.md)
Generate `docs/SYSTEM_MAP.md` + glossary, write `init.sh` (sync + verify) **and `run.sh` (start the app)** using the project's real commands, and capture the **baseline test state** (green or known-failing) before any harness work.
3. **Safe init & seed** → [references/phase-3-init-and-seed.md](references/phase-3-init-and-seed.md)
Run `harness init` without clobbering existing files, merge the harness operating rules into the project's existing CLAUDE.md/AGENTS.md, seed the backlog, then hand off to [Route 2](../workflow-intake/references/route-2-us-execution.md).
## Hard gates
- **Never clobber**: do not overwrite an existing `CLAUDE.md`, `AGENTS.md`, `init.sh`, `.gitignore`, or any source file. Where the project already has one, **merge** the harness section in and tell the user what changed. `harness init` itself skips existing files and reports them; honor that.
- **Ask-user before touching the repo**: confirm the classification (this is a legacy onboard, not a rewrite) and the discovered build/test commands before writing anything.
- **Baseline honesty**: record the real starting test state. If the project's tests already fail, capture that as a *known-failing baseline* — do not "fix" it as part of onboarding, and do not block onboarding on it. Subsequent feature work must not be stacked on a baseline the user has not acknowledged.
- **Durable artifacts**: the survey report and `SYSTEM_MAP.md` go to committed locations (`docs/`), not transient `.harness/reports/` (see the task-state and durable-vs-transient rules in `AGENTS.md` and [task-state-convention](../../resources/task-state-convention.md)).
- After seeding the backlog, **STOP** — hand control back so the user picks one feature (WIP = 1). Do not start implementation in the onboarding session.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!