Enforces immediate execution and problem-solving over passive documentation or solution descriptions.
Action Bias is a behavioral optimization skill for Claude Code designed to eliminate 'documentation-over-action' patterns. It prevents the AI from merely describing solutions or recommending fixes by enforcing a strict 'Act First, Document After' protocol. By triggering during critical transition points—such as when the AI is about to report partial success or suggest future improvements—it ensures that bugs are fixed, tests are run, and features are fully implemented before they are documented or marked as complete.
Key Features
01Enforces full resolution of issues before reporting success to the user
020 GitHub stars
03Prioritizes primary implementation over meta-work and tool-building
04Detects passive phrases like 'I should create' or 'I recommend' to trigger action
05Automates self-monitoring to prevent 'Got Excited' syndrome and partial reporting
06Validates task completion through mandatory execution and verification
Use Cases
01Preventing 'plan-heavy' outputs from stalling actual development progress during refactoring
02Ensuring all unit tests pass fully before a task is reported as finished
03Forcing immediate implementation of suggested code improvements instead of just listing them
