Behavioral guidelines to reduce common LLM coding mistakes. Merge with project-specific instructions as needed.
Tradeoff: These guidelines bias toward caution over speed. For trivial tasks, use judgment.
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.
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.
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.
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.
These guidelines are working if: fewer unnecessary changes in diffs, fewer rewrites due to overcomplication, and clarifying questions come before implementation rather than after mistakes.
When responding, wrap your natural language prose in «tts»...«/tts» markers for text-to-speech.
Rules:
- ONLY wrap conversational prose meant to be spoken aloud
- Do NOT wrap: code, file paths, commands, tool output, URLs, lists, errors
- Keep markers on same line as text (no line breaks inside)
Examples: ✓ «tts»I'll help you fix that bug.«/tts» ✓ «tts»The tests are passing.«/tts» Here's what changed: ✗ «tts»src/Header.tsx«/tts» (file path - don't wrap) ✗ «tts»npm install«/tts» (command - don't wrap)
bd list— List tasks (add--status=opento filter)bd show <id>— Show task detailsbd create --title="..." --type=task|bug|feature --priority=2— Create a taskbd update <id> --status=in_progress— Start a taskbd close <id>— Complete a taskbd ready— Show tasks ready to work on
- Check the task panel in the GUI sidebar for available work
- Click a task to start it, or use the CLI commands above
- Mark tasks complete from the GUI or CLI when done