Stop That Shit: Curb Agent Over-Engineering

Where this fits

You ask Codex to fix a one-line bug. It comes back with a new utils/ directory, a subagent for validation, and SHA-256 checksums for every file. This skill—Stop That Shit—is a configuration layer that tells agents to keep changes minimal and focused. It solves the real problem of AI coding assistants turning trivial tasks into architectural refactors, wasting time and review effort.

Why agents benefit

• Scope enforcement: Agents read the skill’s rules and limit changes to the exact files needed for the task.
• File-count control: It prevents splitting a single function into multiple modules when the original is fine.
• Dependency discipline: Agents avoid adding new packages or libraries unless explicitly requested.
• Hash suppression: It stops agents from generating checksums or verification hashes that add noise without value.
• Subagent restraint: It discourages spawning subagents for tasks that can be done in the main loop, reducing overhead.

Practical scenarios

Fixing a one-line bug

You ask Claude Code to correct a typo in a variable name. Without the skill, it might create a constants.js file and refactor the whole module. With Stop That Shit, the agent edits the single line and runs tests—done.

Adding a new API endpoint

You need to add a POST /users route. The agent with this skill will modify the existing router and controller, skipping the temptation to introduce a service layer or a custom error-handling middleware unless you ask for it.

Refactoring a small function

You want to rename a function for clarity. The skill ensures the agent only updates the definition and call sites, not the entire codebase’s style or structure.

Add it to your agent workflow

Install it with:

npx skills add lennney/stop-that-shit

Then reference it in your agent’s instructions (e.g., AGENTS.md or a project-level config):

# Coding Rules
- Apply the "Stop That Shit" skill before making any changes.
- If the task is a small fix, do not create new files, add dependencies, or generate checksums.

For a full list of similar skills for coding agents, see the OpenAgentSkill listing at https://www.openagentskill.com/skills/lennney-stop-that-shit.

Compare before adopting

Before adding this skill, compare it with alternatives like minimal-diff or custom agent prompts. Check:

• Quality signals: Does the skill have recent commits, tests, and clear documentation?
• Maintenance freshness: Is the author actively responding to issues or PRs?
• Workflow fit: Does it conflict with your team’s existing conventions or agent settings?
• Alternatives: Sometimes a simple .claude-code rule or a one-line system prompt achieves the same effect.

Why it is worth tracking

With 141 stars and an MIT license, this skill shows early community traction. Its narrow focus—preventing over-engineering—addresses a pain point many developers hit daily. Watch for updates that expand compatibility to other agents or add stricter scope rules. Evaluate it when you notice your agents consistently producing more code than necessary; a lightweight guardrail like this can save hours of review time.