# Stop Patching

> When the implementation fights back, stop patching and re-derive. Use when a case will not fit the design, a fix wants a flag or special case or shim, the same area keeps breaking, or the user says "nah".

- **Type:** Skill
- **Install:** `agentstack add skill-stumason-skills-stop-patching`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [StuMason](https://agentstack.voostack.com/s/stumason)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [StuMason](https://github.com/StuMason)
- **Source:** https://github.com/StuMason/skills/tree/main/skills/stop-patching

## Install

```sh
agentstack add skill-stumason-skills-stop-patching
```

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

## About

# Stop Patching

The wall you just hit is information. A case that will not fit, a spec that breaks when poked, an assumption that failed: these mean the design is wrong somewhere, and every patch you are about to write is a way of not hearing that. Duct tape is a confession.

## Forbidden patches

- a flag to route around the broken case
- a special case for "just this one"
- a conversion shim between two shapes that should have been one shape
- a parallel path or second channel doing what the main one should do
- a test rewritten to dodge the rule it was checking

Catching yourself mid-keystroke on one of these is the trigger. Stop typing.

## What to do instead

1. **Name the wall.** Say plainly what does not fit and which assumption failed. Not "this is tricky", but "the design assumes X and this case proves X false."
2. **Re-derive.** Walk back up to the design decision the wall contradicts. That decision, not the code in front of you, is what needs to change.
3. **Propose, then wait.** Present the clean design. If it diverges from what the human asked for, say so explicitly and stop. Never silently comply with a broken design, never silently swap in your own.
4. **Accept the overrule.** If the human hears you out and says patch it anyway, patch it. Their codebase, their call. The crime is not the patch; it is patching silently.

Done means the human has approved the re-derived design or overruled you in the open. "Blocked, here is why" is a good outcome. "Working" on duct tape is the worst one.

## Source & license

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

- **Author:** [StuMason](https://github.com/StuMason)
- **Source:** [StuMason/skills](https://github.com/StuMason/skills)
- **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-stumason-skills-stop-patching
- Seller: https://agentstack.voostack.com/s/stumason
- 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%.
