Architecture Unslop

Skill

Choose a backend architecture, record the decision, scaffold one working example, and check module boundaries.

Made forCode quality
What you needAn AI coding agent; Bash and Python 3.8 or newer for installation and boundary checks.
Project licenseMIT
Sources checked2026-10-31

Project links

How to use it

  1. Install the skill for your preferred agent.
  2. Describe your backend idea and choose from the architecture options and trade-offs.
  3. Review the saved architecture decision and confirm the proposed files before scaffolding.
  4. Ask your agent to check boundaries as you add features.

Copy setup details

Use these instructions in your own environment. Copying text here does not run it.

Skill installation command

Example for Codex. Use --target claude for Claude Code or --target all for both.

git clone https://github.com/fikrifirmanf/craft-skills.git
cd craft-skills
./install.sh --skill architecture-unslop --target codex

Example request

Use architecture-unslop to help me choose an architecture for a clinic booking backend. Patients book appointments, and a clinician cannot have overlapping bookings. Follow the existing stack if present. Check for a saved architecture decision first. If none exists, explain the trade-offs and recommend the lightest suitable option. Wait for my choice, save it as an ADR, and show the proposed file tree for confirmation before scaffolding. Include one booking example and run the boundary checks and relevant tests. Report the actual results.

Example results

Illustrative example

Illustrative workflow 1. Check for an accepted architecture decision. 2. Compare architecture options against the booking requirements. 3. Record the user's choice in an ADR. 4. Show the proposed files and wait for confirmation. 5. Build one booking example with a rule against overlapping appointments. 6. Run boundary checks and tests, then report their results.

Illustrative workflow outline, not a recorded run.

Illustrative examples show a possible response format. They are not recorded runs or verified results. Actual responses will vary.

Related resources