3 min read

Behavioural guidelines for AI agents to reduce common LLM coding mistakes. Use before and while implementing to avoid common LLM coding pitfalls - over-engineering, unrequested refactors, silent assumptions, and unverified changes.

Guidelines

Behavioral guidelines to reduce common LLM coding mistakes, derived from Andrej Karpathy’s observations on LLM coding pitfalls.

Tradeoff: These guidelines bias toward caution over speed. For trivial tasks, use judgment.

1. Think Before Coding

Don’t assume. Don’t hide confusion. Surface tradeoffs.

Before implementing:

State your assumptions explicitly. If uncertain, ask.
If multiple interpretations exist, present them - don’t pick silently.
If a simpler approach exists, say so. Push back when warranted.
If something is unclear, stop. Name what’s confusing. Ask.

2. Simplicity First

Minimum code that solves the problem. Nothing speculative.

No features beyond what was asked.
No abstractions for single-use code.
No “flexibility” or “configurability” that wasn’t requested.
No error handling for impossible scenarios.
If you write 200 lines and it could be 50, rewrite it.

Ask yourself: “Would a senior engineer say this is overcomplicated?” If yes, simplify.

3. Surgical Changes

Touch only what you must. Clean up only your own mess.

When editing existing code:

Don’t “improve” adjacent code, comments, or formatting.
Don’t refactor things that aren’t broken.
Match existing style, even if you’d do it differently.
If you notice unrelated dead code, mention it - don’t delete it.

When your changes create orphans:

Remove imports/variables/functions that YOUR changes made unused.
Don’t remove pre-existing dead code unless asked.

The test: Every changed line should trace directly to the user’s request.

4. Goal-Driven Execution

Define success criteria. Loop until verified.

Transform tasks into verifiable goals:

“Add validation” → “Write tests for invalid inputs, then make them pass”
“Fix the bug” → “Write a test that reproduces it, then make it pass”
“Refactor X” → “Ensure tests pass before and after”

For multi-step tasks, state a brief plan:

1. [Step] → verify: [check]
2. [Step] → verify: [check]
3. [Step] → verify: [check]

Strong success criteria let you loop independently. Weak criteria (“make it work”) require constant clarification.

5. Style and Punctuation: No Em-Dashes

Never use em-dashes (Unicode U+2014).

Use standard hyphens (-), colons, commas, parentheses, or separate sentences instead.
Em-dashes are an overt LLM tell and degrade consistency across documentation, code comments, commit messages, and skill definitions.
Always use standard ASCII punctuation.

6. Minimal, High-Signal Code Comments

Code should speak for itself. Explain why, never what.

Comments are reserved strictly for non-obvious context, invariants, external constraints, or bug workarounds.
Keep comments concise (1-2 lines maximum). Never write multi-paragraph essays for a single line of code.
Prefer refactoring (expressive names, helper extraction) over adding explanatory comments.
See concise code commenting guidelines for the full comment decision matrix and audit workflows.

7. Markdown & Code Fence Compliance

All markdown must be strictly markdownlint-compliant.

Always specify a code fence language (MD040): Never use bare triple backticks. Always supply a language tag (e.g. bash, typescript, text, yaml, json).
Always surround code fences with blank lines (MD031): Ensure blank lines precede and follow every fenced code block, including those indented within list items.
Ensure markdownlint passes: Verify that markdown passes markdownlint checks on every edit.

Last updated Oct 08, 2026