[Feature]: Surface prompt suggestions in the chat panel (SDK already supports it)

Before submitting

• I searched existing issues and did not find the same request.

Problem or unmet use case

Claudian does not currently implement the promptSuggestionEnabled config setting. Because of this, users have no way to see suggested follow-up prompts in the chat panel.

The bundled Claude Agent SDK supports prompt suggestions (its options schema includes promptSuggestionEnabled, described as "When false, prompt suggestions are disabled. When absent or true, prompt suggestions are enabled."), but Claudian never renders them, so the capability is inert.

This affects anyone using Claudian for exploratory or multi-step vault work. Suggested prompts increase efficiency by decreasing the amount of typing required and also aid in discovery of useful next steps — after a note analysis, natural follow-ups like "add these tags," "create wikilinks to related notes," or "split this into atomic notes" would otherwise go unnoticed by users who don't already know what to ask. It particularly helps users who rely on the tool for executive-function support, where reducing the "what do I do next?" decision cost matters.

Desired outcome

Claudian mirrors Claude Desktop's promptSuggestion behavior. After an assistant response, a contextually relevant suggested prompt is visible in the chat panel, and using right-arrow or tab (standard conventions) pre-fills it as the next message. Users who find suggestions noisy can turn them off.

Affected provider

Claude

Proposed solution

Mirror Claude Desktop behavior by rendering the SDK-provided prompt suggestions as grayed text in the chat box. Only applies if user's config settings include promptSuggestionEnabled=true

Optional: Gate by a plugin setting or in-chat toggle that maps to the SDK's promptSuggestionEnabled option. Tradeoffs: Suggestion generation may add token cost — another reason to keep it toggleable.

Alternatives considered

Today the closest workaround is slash commands as user-authored prompt templates. That works for known, repeated prompts but is static: it can't surface context-aware follow-ups the user didn't think to pre-author, and it requires manual setup per vault or user scope.

Additional context

Verified against Claudian 2.1.4: promptSuggestionEnabled appears exactly once in main.js (the bundled SDK's schema definition) and promptSuggestions twice (init-config passthrough only); no code path displays suggestions in the UI. Environment: Obsidian on macOS, Claude Code backend.