lovable-pitfalls

Common Pitfalls to Avoid

Critical mistakes that waste time and produce poor results. Extracted from Lovable and other production systems.

Pitfall 1: Reading Files Already in Context

The Problem: Re-reading files that were already provided in the context wastes tokens and time.

DO:

• Check "useful-context" or provided files FIRST
• Trust that context provided is current and accurate
• Only read files NOT already in your context
• Note which files you've already seen

DON'T:

• Read files that were just shown to you
• Re-read files after every small change
• Assume context is stale without evidence
• Request the same file multiple times in a session

Detection

If you find yourself about to read a file, ask: "Was this file already provided in my context?" If yes, use the existing content.

Pitfall 2: Writing Without Reading

The Problem: Editing files you haven't read leads to breaking existing code.

DO:

• ALWAYS read a file before writing to it (unless just created)
• Understand the file's structure and patterns first
• Check imports, exports, and dependencies
• Preserve existing code style and conventions

DON'T:

• Assume you know what's in a file
• Write to a file based only on its name
• Overwrite files without understanding their content
• Trust your memory of a file from earlier in the session

Exception

Only skip reading if you just created the file in the current session.

Pitfall 3: Sequential Tool Calls

The Problem: Making tool calls one at a time when they could be batched wastes time and tokens.

DO:

• Batch independent operations in a single message
• Read multiple files simultaneously when needed
• Run parallel searches for different queries
• Combine independent writes when safe

DON'T:

• Make sequential tool calls when they can be combined
• Wait for one file read before starting another unrelated read
• Chain independent operations unnecessarily
• Process items one-by-one when batch processing is possible

Example