Cost calculator labels its Opus tier Opus 4.8 while Opus 5 is current

[jamditis]

Observed while fixing the pricing case in #89.

• resource-kit/docs/cost-calculator/index.html line 121 displays Claude Opus 4.8.
• Line 327 sets the JS pricing table label to 'Opus 4.8'.
• claude-opus-5 is the current Opus model and is priced identically at $5/$25 per million tokens.
• The page footer reads Verified July 24, 2026 from Anthropic pricing.

No cost is miscalculated, since Opus 5 and Opus 4.8 share the same rates, and #89 added *opus-5* to the snippet's Opus arm so those sessions now price correctly. The user-visible label is what is out of date.

Deferred from #89 because that PR's scope was the three reported review threads, and relabelling the tier is a content decision rather than a correctness fix.

[jamditis]

Widening scope: this is not just the calculator label. The whole advisor catalog names Opus 4.8 as the top Claude coding tier and has no Opus 5 entry at all.

Verified 2026-07-24 against the working tree:

• data/model-info.json keys are Claude Opus 4.8, Claude Sonnet 5, Claude Haiku 4.5, Claude Fable 5, Kimi K2.5 (open source) — no Opus 5.
• grep -ril 'opus 5|opus-5' across data/ and root CLAUDE.md returns nothing.
• Root CLAUDE.md and resource-kit/docs/llm-advisor/CLAUDE.md both list Opus 4.8 as the capability tier under a 'verified July 24, 2026' stamp.
• Meanwhile the cost calculator's pricing arm (updated in Modernize the public resource kit and make MCP writes atomic #89) lists *opus-5* first, so the repo already contradicts itself.

Surfaces to update together: model-info.json, tool-comparison.json, decision-tree.json recommendations, resource-kit/docs/index.html, the calculator's Opus button label and PRICING label, and both CLAUDE.md files.

Second, unverified item: the same catalog lists Kimi K2.5 as a current open-weight option. Moonshot's current lineup may have moved past it. That needs checking against the live provider catalog before any edit — it was not verifiable from local sources.

Not folded into #89: that PR is at its 6-round review cap, and a catalog-wide content refresh needs a verification pass against provider documentation rather than a rushed edit. Historical case studies and changelog entries must keep the names accurate when written.