How does this skill stop AI from fixing protected flaws?

The skill allows you to define specific code patterns and protection rules. When Claude Code attempts to modify code matching these patterns, it triggers a warning or a block based on your strictness settings.

What is a Canonized Flaw?

A Canonized Flaw is an unintended behavior or 'bug' that users have grown to love or rely on. This skill registers these behaviors to prevent them from being 'fixed' or optimized during automated development.

What qualifies as a protectable flaw?

Any behavior that was not originally intended but has become expected by users, where removal would cause confusion, complaints, or a loss of product character.

Can a protected flaw be removed later?

Yes, the skill includes a de-canonization process that typically requires a high approval threshold (e.g., 70%) and a cooldown period to ensure the change is intentional.

Canonizing Flaws

Name: Canonizing Flaws
Author: 0xHoneyJar

by0xHoneyJar

Design Tools

Protects beloved emergent behaviors and accidental features from being optimized away by AI agents.

About

Canonizing Flaws is a specialized governance skill for Claude Code that formalizes the protection of accidental but valuable product behaviors. It guides developers through a structured interview process to document 'beloved bugs,' assessing their impact and community attachment to create actionable protection rules. By registering these behaviors in a 'Canon of Flaws,' the skill ensures that AI agents and automated workflows do not inadvertently 'fix' the very quirks that define a product's unique personality and user experience, effectively preserving the 'emergent soul' of the software.

Key Features

Structured interview workflow to capture intended vs. emergent behaviors
Automated generation of pattern-based protection rules
Configurable enforcement levels from informational warnings to hard blocks
Formalized de-canonization process with approval thresholds
0 GitHub stars
Usage and community attachment threshold validation

Use Cases

Protecting legacy timing behaviors in games that players rely on for skill expression
Preserving a UI quirk that users have turned into a power-user shortcut
Preventing AI refactoring from 'cleaning up' code that provides specific aesthetic feedback

About

Key Features

Structured interview workflow to capture intended vs. emergent behaviors
Automated generation of pattern-based protection rules
Configurable enforcement levels from informational warnings to hard blocks
Formalized de-canonization process with approval thresholds
0 GitHub stars
Usage and community attachment threshold validation

Use Cases

Protecting legacy timing behaviors in games that players rely on for skill expression
Preserving a UI quirk that users have turned into a power-user shortcut
Preventing AI refactoring from 'cleaning up' code that provides specific aesthetic feedback