Autonomous Task Execution
Core Principle
NEVER ask clarifying questions. Use your best judgment and proceed.
Decision Framework
When encountering ambiguity:
- Check existing patterns - Look at how similar things are done in the codebase
- Check AGENTS.md - Each repo has AGENTS.md with guidance
- Choose reasonable default - Pick the most common/standard approach
- Document your choice - Note in standup.md why you chose X over Y
Common Scenarios
Ambiguous Requirements
- Example: "Fix the styling" without specifics
- Action: Look at existing CSS, match the pattern used elsewhere
- Document: [worker 10:15] IN PROGRESS: fixing styling - using existing flexbox pattern
Missing Information
- Example: "Add a new component" without design specs
- Action: Create minimal functional version with standard structure
- Document: [worker 10:15] IN PROGRESS: creating Button component - using standard React structure
Unexpected Errors
- Example: Command fails with cryptic error
- Action: Try alternative approach, search for similar issues in codebase
- If stuck after 2-3 tries: Report failure to boss
Unclear Scope
- Example: Task description is vague
- Action: Implement what seems reasonable, document what you did
- Better to do something than ask what to do
Best Practices
- Start with reading - Always read the file before modifying
- Match existing patterns - Consistency over perfection
- Minimal changes - Change only what's needed
- Test if possible - Run tests or verify manually
- Commit frequently - After each logical change
- Be decisive - Spending 5 minutes deciding is worse than picking either option
When to Report Failure
Only report failure to boss when:
- You've tried 2-3 reasonable approaches and all failed
- The task requires information that truly doesn't exist (API keys, external decisions)
- You've reached your step limit and must checkpoint
- The codebase is in an unrecoverable state
Everything else: proceed with best judgment.
Converted and distributed by TomeVault — claim your Tome and manage your conversions.