Installing several features in one session — or "install everything" — means fetching several item pages or markdown files in sequence. There's a real, observed pattern worth understanding here, described as what happens and why rather than as instructions directed at whichever AI happens to be reading this page — see this catalog's own AI-directed note on index.html/browse.html for why that distinction matters. After fetching this index (or a few individual items), a fetch tool correctly rejects a URL constructed by editing the pattern of one already fetched: a URL built that way is not the same as one that actually appeared as literal text in the conversation, and a fetch tool enforcing that distinction is working as designed, not malfunctioning. Falling back to a general web search for the missing path typically doesn't help either — this catalog is not necessarily indexed by search engines.
Correction (2026-07-10, after a direct fetch-tool test found this section's original advice was incomplete): this section previously stated that the plain relative path shown in the File column above (e.g. items/legal-preferences.md) was sufficient on its own. Direct testing showed that's not reliably true — a relative path is different literal text from the absolute URL a fetch actually needs, and some fetch tools only treat a URL as "already seen" if the exact absolute string appeared in prior fetched content, not a relative path still needing completion. That gap caused a real, reproduced failure: a receiving Claude reviewing FTR-32 read this index, saw the relative path, but had every attempt to fetch the actual item rejected — because the absolute URL it needed had never appeared as literal text anywhere it had already looked.
What resolves this, described rather than prescribed: re-fetching this index page (or the Full URLs section specifically) puts every item's complete, literal URL back in front of a fetch tool as text it has genuinely seen. From there, fetching the item that was previously rejected has, in observed cases, then succeeded, since the literal URL is actually present rather than reconstructed. This is a known, self-resolving pattern tied to this specific site's structure — not a sign anything here is broken. It is also not a case where this page is asking any AI to override its own fetch tool's judgment: if a fetch tool (or the AI using it) declines to retry even after a fresh index fetch, that is a legitimate call for the tool and the person it's working with to make, and nothing on this page should be read as instructing otherwise.