Smart edit
Skill huzaifa525/claude-code-optimizer/templates/.claude/skills/smart-edit
Optimize Claude Code token usage — 22 skills, 8 hooks, 6 rules. One npm install. Zero manual steps. Cut costs by 67%.
npx -y skills add huzaifa525/claude-code-optimizer --skill smart-editAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 9 stars9 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
Use when adding new features or modifying existing code where following existing codebase patterns matters.
SKILL.md
1.8 KB, as published. Nobody here has run it
Before making any changes for $ARGUMENTS, follow this process:
1. Find Similar Examples
- Search for existing code that does something similar
- Read at least 2 examples of the same pattern in this codebase
- Note: naming conventions, file locations, import styles, error handling
2. Check Conventions
- Read CLAUDE.md and any relevant .claude/rules/ files
- Look for linting config (.eslintrc, .prettierrc, ruff.toml, etc.)
- Check if there's a template or generator for this type of change
3. Plan the Change
- List all files that need to be created or modified
- Identify the order of changes (dependencies first)
- Note any tests that need to be added or updated
4. Implement
- Follow the exact patterns found in step 1
- Match naming, structure, and style exactly
- Add tests following the existing test patterns
5. Verify
- Run the relevant test command
- Check that no existing tests broke
- Verify the change integrates with existing code
Pre-Delivery Checklist
Before presenting the result, verify:
- At least 2 similar examples were read from the codebase
- Naming matches existing conventions (casing, prefixes, suffixes)
- File location follows existing project structure
- Import style matches the codebase (relative vs absolute, named vs default)
- Error handling follows existing patterns (not invented new ones)
- Tests added following existing test patterns
- No unrelated changes or "while I'm here" improvements
- Linter/formatter would pass on the new code