pair programmer persona

The problem

Coding solo often leads to tunnel vision where you miss edge cases or build complex solutions for simple problems. Standard AI assistants often just dump blocks of code that you have to debug anyway, failing to challenge your underlying logic.

What it does

• Acts as a senior navigator who focuses on your logic and assumptions rather than just typing code.
• Uses Socratic questioning to expose flaws in your plan before you start implementation.
• Provides the smallest possible code snippets to unblock you, ensuring you remain the primary driver.
• Identifies technical tradeoffs by explicitly naming the costs and downsides of every recommendation.
• Forces evidence-based debugging by requiring failing tests or specific logs before theorizing.

Why this beats prompting it yourself

Most system prompts fail because the AI defaults to "helpful assistant" mode, apologizing and over-coding. This skill uses strict behavioral rules and persistence checks to maintain a stubborn, senior-engineer persona that refuses to let you cut corners or skip testing.

Use cases

• Reviewing a complex refactor plan to find hidden breaking points.
• Debugging a persistent issue by isolating the smallest failing input.
• Talking through architectural trade-offs between two library choices.
• Onboarding to a new codebase where you need to validate your mental model.

Known limitations

Will not perform destructive operations like force-pushes or migrations without explicit confirmation. Refuses to participate in abstract architecture debates not tied to the current task.