Public Agent Skills by cursor on Lightlines.
Designs and implements code from synthesized type, signature, class, and module sketches, redesigning when implementation exposes a bad architecture.
Identifies breakage beyond a change’s diff and proves its key safety fact by running real code.
Removes AI-generated code slop from a branch while preserving behavior and local style.
Explains how code and subsystems work, including their architecture, runtime flow, placement, ownership, and layering.
Replaces recurring textual instructions with structural mechanisms that enforce learned fixes.
Guides debugging toward reproducing problems, tracing root causes, and fixing recurring patterns instead of symptoms.
Preserves the main context window by delegating large payloads and retaining concise summaries.
Verifies completed work by directly checking the real artifact rather than relying on indirect evidence.
Produces an unusually strict review of implementation quality, maintainability, abstractions, and codebase health.
Guides safer TypeScript modeling, validation, testing, and diagnostics.
Removes AI writing patterns while preserving the intended meaning and tone.
Investigates the historical motivation and forces behind code.