The agent assembled a reasonable explanation of ShadowClaw by combining minimal content from the homepage with substantial technical details from the GitHub repository. The xt-ml.github.io site itself is thin and deployment-focused; most differentiator details and architecture facts came from the linked GitHub repo, requiring the agent to navigate off-site to answer the comparative question.

• ›The homepage (steps [0–1]) contained only high-level project description; no pricing, licensing, or competitor positioning was present on xt-ml.github.io itself.
• ›The agent correctly identified the need to follow the GitHub link to find AGPLv3 licensing and technical differentiators, showing sound off-site navigation to fill gaps the site left open.
• ›GitHub README (step [3]) was the primary content source for architecture details (Web Workers, OPFS, AES-256 encryption, 50+ tools), making it essential to the final answer despite not being the official domain.
• ›The web search (step [4]) returned no relevant results for ShadowClaw itself, only OpenClaw alternatives, revealing limited market visibility or indexing of ShadowClaw—a legitimate confection the agent correctly noted.
• ›No use-case examples, real-world deployment stories, or explicit competitor comparisons appeared on either the site or GitHub README, forcing the agent to rely on inferred positioning from feature lists rather than stated market positioning.