Feature Request: Extensible prompt suggestions (cursor autocomplete)

Summary

Allow users or tools to inject custom suggested prompts into the cursor suggestion space (the autocomplete suggestions that appear when the cursor is in the input field).

Motivation

Claude Code currently auto-generates suggested next prompts based on conversation context (e.g., "commit this", "what should we work on next"). These are useful but not extensible.

For projects with structured workflows (roadmaps, task queues, session protocols), the most valuable next action is often domain-specific and computable -- but there's no way to surface it where the user's attention already is: the input prompt suggestions.

Proposed Behavior

A mechanism (hooks, config, or API) that allows injecting custom suggestions into the prompt autocomplete. For example:

• A prompt-suggestions hook that returns a list of suggested prompts
• A config file (e.g., .claude/suggestions.json) that feeds static or dynamic suggestions
• A post-tool-call hook that can append suggestions based on what just happened

Use Case

Our project has bootstrap scripts, attention queues, and roadmaps that know what the operator should do next. We want to surface actions like "Review pending alignment votes", "Continue Security Roadmap Milestone 2.3", etc. directly in the suggestion space rather than only in terminal output that scrolls away.

This would turn the suggestion space from a passive inference into an active, user-controllable surface -- consistent with the extensibility philosophy of hooks and CLAUDE.md.