Skip to content

Quick start

This walkthrough takes you from a blank repo to your first /genesis invocation, end to end. You will type one prompt, read the proposed layout, and accept (or reject) the architect handoff before any file is written.

Terminal window
npx skills add danielmeppiel/genesis

Zero global install. Works with Claude Code, Cursor, Codex, OpenCode, GitHub Copilot, and 41+ more agents — see skills.sh.

Already using apm (manifest + lockfile)?

Terminal window
apm install danielmeppiel/genesis

For full install paths, post-install verification, and harness detection, see Install.

In your agent (Claude Code, Cursor, GitHub Copilot, OpenCode, Codex), type /genesis followed by what you want designed:

/genesis I want every PR on my repo reviewed for missing tests, undocumented
public API, and unsafe migrations -- aggregated into one comment. Never
approve, never auto-merge.

You will get a named pattern, an execution diagram, an acceptance test, and a written plan — before any file is touched.

Before writing a single file, Genesis proposes a layout grounded in the catalogue. For the prompt above, expect something like:

.github/skills/pr-review/
|-- SKILL.md # entrypoint, 8-step contract
|-- agents/
| |-- pr-tests-lens.agent.md # missing-tests reviewer (fresh context)
| |-- pr-docs-lens.agent.md # public-API doc reviewer
| |-- pr-migration-lens.agent.md # unsafe-migration reviewer
| `-- pr-synthesizer.agent.md # arbiter, dissent-weighted
|-- rules/
| `-- review-output.md # auto-attached output schema
|-- assets/
| |-- severity-rubric.md # acceptance gate
| `-- findings.template.md # plan persistence shape
`-- triggers/
`-- on-pull-request.yml # event binding

Then it justifies each piece against the catalogue. Patterns are cited by name (A1 Panel, B1 Fan-Out + Synthesizer, S4 Validation Decorator, B4 Plan Memento, A6 Event-Driven). The architect tells you not just what to build, but why this and not that.

Genesis emits a class diagram of the runtime shape before you write any code:

classDiagram
    class PrReviewSkill {
        +SKILL.md
        +invoke(pr_diff)
    }
    class TestsLens {
        +context: fresh
    }
    class DocsLens {
        +context: fresh
    }
    class MigrationLens {
        +context: fresh
    }
    class Synthesizer {
        +arbitrate(findings) verdict
    }
    class OutputSchemaRule {
        <<auto-attached rule>>
    }
    class FindingsMemento {
        <<plan persistence>>
    }
    PrReviewSkill o-- TestsLens : spawns
    PrReviewSkill o-- DocsLens : spawns
    PrReviewSkill o-- MigrationLens : spawns
    PrReviewSkill o-- Synthesizer : spawns
    TestsLens ..> OutputSchemaRule : enforces
    DocsLens ..> OutputSchemaRule : enforces
    MigrationLens ..> OutputSchemaRule : enforces
    Synthesizer ..> OutputSchemaRule : enforces
    Synthesizer ..> FindingsMemento : writes

The file you eventually author is the easy part. The composition above is what was missing.

The architect step ends with a handoff packet: diagrams + interface sketch + declared targets + module composition table + todos. A separate coding step (your agent’s normal coder loop) then turns the artifacts into actual files. Diagrams come first; files come second.

If the proposed layout is wrong, reject it now — it is cheap. Once files exist, restructuring is expensive.

  • Different prompt, different architecture. Three cold-load runs of the same skill yield three materially different output shapes. See the examples gallery (coming soon) for worked walkthroughs of release-notes, advisory PR review, and verdict-emitting PR review.
  • Drill into the substrate. The output above cites primitives by name. Start with Primitives for the six concepts every harness implements.
  • Wire it for your harness. /genesis resolves differently on each runtime. See Harness setup for file paths, frontmatter dialects, and known limitations.