# Architect

> Sketch types, signatures, and module structure before code, then stay in the loop while implementation fills in. Use for architect this, design this, or non-trivial work where jumping to code would lock in the wrong shape.

- Skill: `shipshitdev/architect` (Agent Skill, multi-file: 11 files)
- Install (CLI): `npx skillmds add shipshitdev/architect`
- Raw SKILL.md: https://api.skillmd.com/api/skills/shipshitdev/architect/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- License: MIT
- Author: shipshitdev (https://skillmd.com/u/shipshitdev)
- Updated: 2026-09-09
- Page: https://skillmd.com/skills/shipshitdev/architect

---


# Architect

Design before implementing. Sketch types, function signatures, class
shapes, and module boundaries with `not implemented` bodies. Synthesize
across parallel candidates, then fill in code against the chosen sketch.

Companion to `codebase-design`, which owns deep-module vocabulary. This
skill owns the pre-implementation sketch loop.

## Authorized Scope

Apply this engine only within the user's requested task and existing explicit
authorization. Loading or delegating to it grants no additional authority.
Preserve report-only restrictions and the caller's target, host, provider, and
cost limits. Existing approval satisfies a gate only for the same actions and
scope; obtain approval before expanding them. Forward these limits to delegates.

## Contract

Inputs:

- A change that crosses a function boundary, or non-trivial work whose
  shape is still unset

Outputs:

- A synthesized design package: caller usage first, then types,
  signatures, module map, and rationale
- Implementation against that sketch, or a scrap-and-redesign

Creates/Modifies:

- Design artifacts and, after agree, production code

External Side Effects:

- None beyond local files unless a later playbook opens a PR

Confirmation Required:

- Opt-in checkpoint when the invoker asks to see the sketch first

Delegates To:

- `how` for grounding, `why` when ownership or layering is the
  constraint, `arena` for candidate sketches, `interrogate` when the
  design is contested

## Phases

Open a todo list with one entry per phase: Ground, Sketch, Agree,
Implement, Scrap.

### A. Ground

Run `how` over every system the new code touches. If the design
redefines ownership or layering, also run `why` on the existing shape.
Skip only when the work is genuinely greenfield.

### B. Sketch

Run `arena` with the design-sketch task. Each candidate produces a
package shaped per
[references/rationale-template.md](references/rationale-template.md)
using [references/runner-prompt.md](references/runner-prompt.md).

Require at least two structurally distinct candidates. Screen every
candidate against
[references/design-red-flags.md](references/design-red-flags.md).
Prefer the design that hides more complexity behind a smaller public
surface.

### C. Agree

Default: proceed to implementation. Opt in to a checkpoint when the
invoker asks to see the sketch first. Human pushback is Phase A
evidence. Re-ground and re-run Phase B.

### D. Implement

Replace `not implemented` bodies. Deviations from the sketch are
signal. Surface them. Do not bolt them on silently.

### E. Scrap

If implementation keeps producing the same shape of friction, throw the
sketch out. Re-run `how` over what was built. Redesign as if the new
constraints were day-one assumptions. Subtract before adding. Return to
Phase B.

## Architect procedure

Read [architect procedure](references/architect-procedure.md) when running this workflow.
Apply the authorized scope and mode of this entry point to every step.
Resolve other skills through this distribution’s active catalog; resolve
resources relative to the installed skill directory.

