Show HN: Reduck GEO, open source Skill to measure and optimize Claude citations

Show HN has listed Reduck GEO, an open source Skill for measuring and optimizing Claude citations. The project lives under the reduck-ai/agents repository on GitHub. On Hacker News the thread sits at 1 point with 0 comments, so the submission is still early in public discussion.

Reduck GEO is framed as a Skill rather than a full application or hosted SaaS. That design fits agent workflows where a discrete capability is loaded into Claude and used to inspect how the model cites sources, then to adjust prompts, retrieval, or content so those citations improve. The open source placement under reduck-ai/agents makes the measurement and optimization logic inspectable and forkable instead of locked behind a vendor API.

For engineers and builders shipping Claude-based products, citation quality is not cosmetic. Weak or opaque citations erode trust in RAG answers, support bots, and research agents. A Skill that both measures and optimizes that behavior gives teams a repeatable loop: score how Claude attributes claims, change the agent or content path, and recheck without inventing a custom eval harness from scratch.

Competitive context is the rising focus on GEO-style visibility inside AI answers, not only classic SEO rank. Most teams either ignore citation behavior or bolt on ad hoc prompts. An open Skill aimed at Claude sits between pure research evals and closed optimization platforms: lower cost of entry, tied to one major model surface, and dependent on community adoption rather than a paid ranking product.

Practical takeaway: clone the reduck-ai/agents tree, wire Reduck GEO into a Claude agent that already answers from your docs or web sources, and run a baseline citation measure before changing content or retrieval. Watch whether HN discussion and repo activity grow past this 1-point, 0-comment show page; that will signal whether builders treat it as a real eval primitive or a one-off Show HN listing.