The agent worked around a 403 access barrier by pivoting entirely to web search, assembling a coherent explanation of TSMC's ESG hub from search result snippets and cited articles. The site itself is not directly navigable, but its content (reports, academy pages, articles) is discoverable and indexed, allowing the agent to infer function and audience from titles, metadata, and brief descriptions rather than from full page content.

• ›Direct access to esg.tsmc.com blocked (403 Forbidden), forcing complete reliance on web search results and link titles. The agent did not attempt to authenticate or use alternative tools despite the fetch error suggesting authentication might help.
• ›Search results reliably surfaced specific article pages and reports (e.g., Sustainability Report PDFs, Academy announcements) by title and URL, allowing the agent to infer core offerings without reading full page bodies. Search snippets contained enough context (e.g., '1.2 million users,' 'first open educational platform') to support claims.
• ›Pricing information absent from both search results and site structure — the agent correctly inferred 'free' from the Academy being described as 'free of charge' and 'open to the general public,' but this is site-content inference, not explicit pricing documentation.
• ›No competitive analysis available — the agent could not find or fetch third-party comparisons or alternative ESG platforms, limiting the 'how it's different' section to TSMC's own claim of being 'industry-first.'
• ›Agent's final caveat about 403 Forbidden is accurate and transparent, but the agent did not explore whether the site requires geographic whitelisting, cookie-based access, or specific user-agent headers — it treated the error as terminal rather than a routing problem.