Improve codebase architecture
Portable Agent Skills for coding agents, including Svelte/SvelteKit and workflow tooling skills.
npx -y skills add spences10/skills --skill improve-codebase-architectureAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 13 stars13 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Find AI-readiness architecture issues and module-deepening refactors. Use when improving codebase architecture, testability, agent navigation, seams, or refactor tickets.
SKILL.md
2.8 KB, as published. Nobody here has run it
Improve Codebase Architecture
Your codebase is the real prompt. Use this skill to find places where agents will get lost, make scattered changes, or accelerate entropy, then turn them into human-approved module-deepening refactors.
Core stance: agents are tactical programmers; the human is the strategic programmer. Do not AFK-rewrite architecture.
Quick Start
- Explore the codebase like a new starter with no memory.
- Return a numbered shortlist of deepening candidates; do not design solutions yet.
- Ask: “Which candidate should we grill?”
- Ground the chosen candidate in code, ask boundary questions, then sketch a deep/gray-box module.
- If requested, create a focused RFC/issue for an implementation agent.
Process
Explore for AI friction. Build the mental map exposed by the filesystem: product/domain areas, obvious public interfaces, tests/feedback loops, and places requiring too much hidden context. The friction you experience is signal.
Find deepening opportunities. Prioritize candidates that improve locality and leverage: duplicated domain rules, parallel implementations with no seam, shallow modules, arbitrary imports, missing boundary tests, and hard-to-discover public APIs.
Present candidates, not solutions. For each candidate include: modules involved, why this is AI friction, proposed deepening move, locality gain, leverage gain, seam/test impact, confidence, and risk. Then ask which candidate to grill.
Grill the chosen candidate. Ground it in concrete files, call sites, invariants, dependencies, and test gaps. Ask what the module owns, what interface callers should learn, what invariants tests must lock down, which dependencies need adapters, and how to migrate safely.
Sketch the module shape. Propose a deep/gray-box module: public interface/types, common usage example, hidden implementation details, seams/adapters, boundary tests, and migration steps. Optimize for progressive disclosure: agents should understand the module from interface/docs before reading internals.
Produce the handoff artifact. If the user wants follow-through, create a focused refactor RFC/issue. Do not implement broad architectural rewrites unless explicitly asked; if implementing, do one narrow module-deepening slice with tests.
Reference
Read REFERENCE.md for the glossary, candidate rubric, design prompts, dependency categories, and RFC template.