furkangonel Simplify Skill
Goal
Run a focused quality audit on the code changed in the current session (or specified files), identify reuse/quality/efficiency issues, then fix them — all without changing external behavior.
Philosophy
- Behavior is sacred: Never change what the code does, only how it does it
- Prefer deletion: The best code is code you don't have to maintain
- Three lenses: Code Reuse (DRY), Code Quality (readability, correctness), Efficiency (performance, bundle size)
Steps
1. Identify changed code
Run git diff --name-only HEAD (or the specified files) to get the scope.
If no git changes and no files specified, ask the user which files to audit.
Success criteria: You have a clear list of files/functions to audit.
2. Launch parallel audit agents
Use spawn_subagent_parallel with three agents running simultaneously:
Agent 1 — Code Reuse (explore) Task: Find duplication in the changed files. Look for:
- Functions/logic duplicated elsewhere in the codebase
- Inline code that should be extracted into utilities
- Repeated patterns that could use existing libraries
- Constants that are hardcoded instead of referenced
Agent 2 — Code Quality (code-reviewer) Task: Identify quality issues in the changed files. Look for:
- Unclear variable/function names
- Missing or incorrect error handling
- Functions doing too many things (SRP violations)
- Commented-out code or dead branches
- Type safety issues (any, implicit any, missing return types)
Agent 3 — Efficiency (performance) Task: Find performance and efficiency issues in the changed files. Look for:
- Unnecessary re-computation inside loops
- Missing memoization for expensive operations
- Synchronous blocking where async would work
- Large imports when only small parts are used
- Memory leaks (event listeners, timers, subscriptions not cleaned up)
Success criteria: All three agents complete and return findings.
3. Triage findings
Collect all findings from the three agents. Categorize each as:
- Fix now — clear improvement, zero risk of behavior change
- Discuss — might be improvement but has trade-offs (share with user)
- Skip — out of scope or not worth the churn
Show the Discuss items to the user before touching them.
Success criteria: User is aware of any non-obvious changes before they're made.
4. Apply fixes
For each "Fix now" item, apply the change using edit_file. After all edits:
- Run any available lint/type-check:
tsc --noEmit,eslint,ruff, etc. - If tests exist, run them:
npm test,pytest, etc.
Success criteria: All checks pass, no behavior change.
5. Report
Summarize what was changed and why, grouped by the three audit lenses. Include before/after snippets for the most significant changes.
Use send_message to present the summary.
Why/Failure Modes
[TODO: Explain the reasoning behind this skill's approach and common failure modes to avoid.]
Standalone vs Supercharged
[TODO: Describe how this skill works on its own vs when combined with other tools/context.]
Cross-References
[TODO: Link to other relevant skills or documentation.]