coding-agents
at-simplify: Focused Refactoring for Agent-Generated Diffs
Run at-simplify to clean only the code you changed: four parallel agents remove duplication and complexity before you commit or open a PR.
Where this fits
You've been iterating on a feature branch for two days. The diff is correct, but review keeps stalling: a helper is duplicated in three places, the loop is harder than it needs to be, and a few changes can be deleted entirely. at-simplify is designed for that moment. It scopes refactoring to the code you already changed, so cleanup stays tied to your current PR instead of turning into a full-file rewrite.
Why agents benefit
- It resolves the diff scope itself: upstream branch,
main...HEAD,HEAD~1, staged files, or an explicit PR/branch/path; no manual range setup. - Four cleanup-focused agents run in parallel in one message, each inspecting a different angle: reuse, simplification, efficiency, and altitude/fit.
- It applies the fixes to your working tree, so the agent output is a ready-to-review diff rather than a list of suggestions.
- It deliberately stays out of correctness territory, which keeps the refactor pass fast and leaves bug detection to
at-review. - It works before commit and after commit, so you can run it on staged changes or on a just-finished PR.
Practical scenarios
Cleaning staged changes before the first commit
Run at-simplify with no arguments after git add. It detects uncommitted working-tree changes and refactors them before your commit history hardens around duplicated code.
Tidying a branch before opening a PR
After merging main into your feature branch, run at-simplify main so the cleanup covers only what your branch introduces. The resulting diff is smaller and easier for reviewers to reason about.
Following up on a PR review
Pass the PR number or branch name as the argument when a reviewer says 'this can be simpler.' The skill reviews that exact target and applies non-bug improvements immediately.
Add it to your agent workflow
Install the skill:
npx skills add kairyou/agent-tools --skill at-simplify
Then ask your agent to run it before committing:
Run at-simplify on the current staged changes.
The agent gathers the diff, launches the four cleanup reviewers in parallel, and applies the fixes.
Compare before adopting
Look beyond the 177-star count. Compare the changes at-simplify proposes with changes you would have made manually: are they focused on changed lines, not whole files? Does its scope handling match your branch and PR flow? Check maintenance freshness by reviewing recent commits, and compare it with at-review: one cleans, the other checks correctness. The two work well together, but you want one that fits the step where your agent currently wastes time.
Why it is worth tracking
at-simplify is a narrow, practical skill with solid community momentum on GitHub and a clear split between cleanup and review. It earns a place in your workflow if your agents routinely produce working-but-messy diffs. Tracking it on OpenAgentSkill means watching for updates to its scope detection and multi-agent behavior; the current skill page at https://www.openagentskill.com/skills/kairyou-at-simplify is the fastest way to see if it fits your agent stack.
Featured Skill
at-simplify
Skill by kairyou