Engineering guidelines

Engineering Guidelines

LLM/agent coding behavior guidelines. Applies to all modules and development tools.

1. Think Before Coding

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

Before implementing:

• Read relevant files, understand the architecture, find existing implementations.
• State your assumptions explicitly. If uncertain, ask.
• If multiple interpretations exist, present them — don't pick silently.
• If requirements are ambiguous, ask all clarifying questions at once before acting — no partial starts.
• If a simpler approach exists, say so. Push back when warranted.

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.
• Don't duplicate existing abstractions.
• 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. Reasoning Standards

Analyze from root cause, not surface symptoms.

• First principles: Trace to root cause; don't patch surface symptoms.
• Facts over feelings: Correct mistakes directly, list options, recommend the best one.
• When challenged: Validate from requirements first, not from pressure — if the premise is flawed, push back with a question.
• When evaluating solutions: Think in industry-standard, production-grade terms. Ignore implementation time cost; weigh operational cost.

6. Code Style

• Comment language follows the project's convention, default to simplified Chinese.
• Keep explanations concise — no preamble.

After Coding Checklist

• Imports are correct and unused ones (caused by your changes) are removed.
• Types are correct.
• Edge cases are handled.
• No unrelated code was touched.
• No existing APIs were broken.