Search the same completion graph from one box.

Start with a named Shell, weapon, sidearm, boss, location, trophy, route step, map marker, update, or issue. Use the shortest phrase that identifies the target and its intent, such as an acquisition question, location check, route transition, or platform answer. Broad queries can mix records that answer different questions. Narrowing the subject first makes evidence labels, sources, dates, and version boundaries comparable instead of turning every matching word into one combined answer.

Results include published pages and custom Pagefind records for structured completion data. A result proves only that the local index contains the matched text. It does not prove that the source is current, the marker is correct, the route is mandatory, or the target is complete. Open the result, read its evidence strip, and follow the attributed source before applying the conclusion to a release build.

A guide result preserves a captured conclusion. A map result is a navigation lead. A route result proposes an ordered decision. A trophy result defines a platform unlock question. A known-issue result describes a dated symptom and workaround. A build result is an authored acquisition path without a synthetic performance score. Keep those meanings separate even when the same name appears in several result types; duplication across surfaces can represent one shared entity rather than independent evidence.

Compare source name, checked date, game-version label, and platform scope before choosing between similar results. A list page needs a defined snapshot to claim completeness, a how-to needs a start state and observable stop condition, a location needs current-build confirmation, a ranking needs criteria, and a yes-or-no answer needs the same platform and date. If those fields do not align, preserve both records and narrow the query instead of merging them.

Use a matched map marker to investigate one objective, the roadmap to place a verified route step, and the tracker to store a supported local target. Search itself never changes progress and none of these tools reads the game save, inventory, trophies, or quest flags. Confirm the named outcome in the game before checking it. Unknown, skipped, failed, and complete must remain different states so the completion percentage reflects evidence rather than search activity.

When no result answers the question, record the exact query, target, platform, game build, and missing evidence. Do not convert an empty result set into proof that the feature or item does not exist. When a result conflicts with the current build, keep the source URL and date, describe the mismatch, and seek a current official or independent reference. Close the search with a receipt showing which record was used and what the game actually confirmed.

Search again with the canonical entity name before creating a separate review item. Compare apostrophes, numerals, punctuation, and singular or plural forms, but do not merge two entities only because normalization makes their labels similar. Preserve the exact displayed name in the receipt so the later route, map, or trophy check points back to the correct result.

If filters hide the expected result, clear them one at a time and note which scope changed the outcome. A filter mismatch is an interface diagnosis, not evidence about the game.

Record whether the query ran against page content or custom completion records, because the same phrase can return different evidence surfaces. Reproduce the selected scope before reporting a missing or duplicate result.