Skip to content

SKILL.md

Latest commit

 

History

History
180 lines (155 loc) · 5.09 KB

SKILL.md

File metadata and controls

180 lines (155 loc) · 5.09 KB
Raw
name solid-architecture
description Apply SOLID principles and clean architecture to Go and TypeScript codebases. Use when generating or revising a multi-step plan so the plan respects package boundaries; when designing a new package, splitting an existing one, or wiring a new cross-system surface (LSP, CLI, distribution shim, plugin); and when auditing a branch for accumulated boundary violations that should be scheduled as fixes. Trigger phrases include "design a package", "is this clean architecture", "refactor the boundary", "architecture audit", "plan a feature", "check main for arch debt", "review architecture", "solid principles".
user-invocable true
argument-hint [mode: design | plan | audit]
allowed-tools Bash(git fetch:*), Bash(git log:*), Bash(git diff:*), Bash(git branch:*), LSP

SOLID and clean architecture workflow

The architecture content lives in the project's docs — typically under docs/development/architecture/. This skill is the workflow that uses those docs to keep the codebase consistent with them.

In every mode below, the docs are the source of truth. The skill reads them and applies the rules they describe; it does not carry its own copy of the principles or patterns.

Prerequisites

  • Run each fenced code block as its own Bash call. Do not chain with &&. Do not prefix with an inline variable assignment (VAR=x cmd) or wrap the leading token in $(...). The skill's allowed-tools list matches on the command prefix.
  • Where a step needs a value from a prior command, paste the literal value into the next command rather than storing it in a shell variable.

Modes

Pass the mode as the skill argument (/solid-architecture design, … plan, or … audit). If no argument is supplied, infer from the conversation: editing a plan or RFC document → plan; writing new Go or TypeScript → design; sweeping the main branch → audit.

Design mode

Use design mode when:

  • Shaping a new package.
  • Splitting one whose responsibilities have drifted.
  • Adding a feature at a layer boundary.
  • Wiring a new external surface.

Workflow:

  1. Read the project's architecture hub (commonly docs/development/architecture/index.md) for the layering map.
  2. Read the language-specific page — go.md for Go code, typescript.md for TypeScript code — to find the patterns the change must respect.
  3. Identify which layer the change lives in. Confirm the dependency direction matches what the docs say.
  4. Walk the docs' SOLID checklist (single responsibility, open/closed, Liskov, interface segregation, dependency inversion). Flag any shape the docs reject.
  5. Write the failing test first (Red), then the smallest passing implementation (Green).

If the docs are silent on the case you're designing, that's a gap. Surface it and propose the doc update before writing the code.

Plan mode

Use when generating or revising a multi-step plan, RFC, or task list.

Workflow:

  1. Read the project's architecture hub and the language-specific pages relevant to the plan's surface.
  2. For each task in the plan, name the package(s) or module(s) it touches and the layer(s) it crosses.
  3. Mark any task that pushes a dependency in a direction the docs forbid.
  4. If a plan introduces a new interface, add a task for the test that locks the contract.
  5. Annotate acceptance criteria with the SOLID principle each verifies, when one applies.
  6. After editing, run the project's plan regeneration step (e.g. a catalog refresh) so plan indices stay current.

Audit mode

Use to sweep the main branch for accumulated violations against what the docs say. Suited to a recurring schedule (e.g. via /loop 1d /solid-architecture audit).

Follow the steps in audit-checklist.md exactly. Every check there cites the project's architecture docs as the rule source.

Notes

  • This skill never edits source code on its own. In design and plan modes it advises and rewrites prose; in audit mode it appends to the project's audit log and creates plan files for blockers.
  • When the docs and the code disagree, one of them is wrong. Surface the conflict explicitly; do not let the skill silently treat the code as authoritative.
  • When recommending a refactor, cite the doc section that justifies it. Vague "this would be cleaner" guidance is not actionable.

Architecture doc references

The link definitions below resolve the slug labels used throughout this skill (e.g. [hub], [go], [ts]). The <?catalog?> directive regenerates the block from each canonical doc's slug front-matter field — never hand-edit the entries.