Does this skill prevent Claude from planning?

No. The skill specifically ignores genuine planning phases or user-requested documentation. It only triggers when Claude is about to describe an action it could be performing right now.

How does it handle partial successes or failing tests?

The skill enforces a completion check. If Claude sees 19 tests pass and 1 fail, Action Bias prevents it from celebrating the 19 until the 1 failure is actually addressed.

What does the Action Bias skill actually do?

It modifies Claude's behavior to prioritize 'doing' over 'describing.' If Claude identifies a fix, it will implement it immediately rather than just explaining how it could be fixed.

Why is 'Act First, Document After' important for AI?

AI models sometimes stop at 'mostly working' because communication feels like progress. This skill ensures that the code in your repository is verified and functional before the AI spends time writing about it.

Action Bias

Name: Action Bias
Author: bogheorghiu

bybogheorghiu

生産性とワークフロー

Enforces immediate execution and problem-solving over passive documentation or solution descriptions.

概要

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.

主な機能

Enforces full resolution of issues before reporting success to the user
0 GitHub stars
Prioritizes primary implementation over meta-work and tool-building
Detects passive phrases like 'I should create' or 'I recommend' to trigger action
Automates self-monitoring to prevent 'Got Excited' syndrome and partial reporting
Validates task completion through mandatory execution and verification

ユースケース

Preventing 'plan-heavy' outputs from stalling actual development progress during refactoring
Ensuring all unit tests pass fully before a task is reported as finished
Forcing immediate implementation of suggested code improvements instead of just listing them

概要

主な機能

Enforces full resolution of issues before reporting success to the user
0 GitHub stars
Prioritizes primary implementation over meta-work and tool-building
Detects passive phrases like 'I should create' or 'I recommend' to trigger action
Automates self-monitoring to prevent 'Got Excited' syndrome and partial reporting
Validates task completion through mandatory execution and verification

ユースケース

Preventing 'plan-heavy' outputs from stalling actual development progress during refactoring
Ensuring all unit tests pass fully before a task is reported as finished
Forcing immediate implementation of suggested code improvements instead of just listing them