# Refactor

> Make the same change across many files safely. Use when renaming a symbol, changing a call signature, moving a pattern, or applying a consistent edit to a whole codebase.

- **Type:** Skill
- **Install:** `agentstack add skill-1aifanatic-super-computer-refactor`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [1aifanatic](https://agentstack.voostack.com/s/1aifanatic)
- **Installs:** 0
- **Category:** [Cloud & Infrastructure](https://agentstack.voostack.com/c/cloud-infrastructure)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [1aifanatic](https://github.com/1aifanatic)
- **Source:** https://github.com/1aifanatic/super-computer/tree/main/skills/refactor

## Install

```sh
agentstack add skill-1aifanatic-super-computer-refactor
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Systematic multi-file edits

The danger in a refactor is not the edit, it is the edit you did not know you needed to make. Find the full blast radius before changing anything.

## Always, in this order

1. **Enumerate.** `grep -rn "oldName" /workspace` — every occurrence, before touching one.
2. **Count.** `grep -rc "oldName" /workspace` per file, so you know when you are done.
3. **Classify.** Definitions, call sites, imports, strings, comments and tests need different treatment. A blind replace corrupts the ones that are not code.
4. **Edit** with `edit_file`, one occurrence at a time with enough surrounding context to be unique.
5. **Verify.** `grep -rn "oldName" /workspace` again. Zero results, or only the ones you deliberately left.

## Use edit_file, not write_file

Rewriting a file to change one line bills the whole file as output tokens and risks losing anything you did not carry across. `edit_file` fails loudly when `old_string` is not unique, which is a feature: that error is telling you the change is ambiguous and needs more context.

When a change genuinely is uniform across a file, `replace_all` is correct — but check the count first so you know what you are agreeing to.

## Partial-word traps

`grep` matches substrings. Renaming `user` will also hit `username`, `userId` and `superuser`. Anchor the pattern:

```
grep -rnw "user" /workspace
grep -rn "\buser\b" /workspace
```

## The limit you must state

You cannot run the tests, the build, or the type checker. A refactor that looks complete is unverified. When you finish, say so plainly — "renamed 14 occurrences across 6 files; not compiled or tested" — rather than implying the change is known good.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [1aifanatic](https://github.com/1aifanatic)
- **Source:** [1aifanatic/super-computer](https://github.com/1aifanatic/super-computer)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-1aifanatic-super-computer-refactor
- Seller: https://agentstack.voostack.com/s/1aifanatic
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.