smart-question-generator

# Smart Question Generator ## Purpose / When to Activate Activate: - As Stage 4 of the `/gaai:bootstrap` pipeline, after `bootstrap-llm-synthesis` produces entries with `clarity: open-question` - The bootstrap orchestrator transforms `open-question` entries from `synthesis_result.entries[]` into the `ambiguity_feed` input (mapping `content` → `topic`, assigning `ambiguity_score` from its own confidence heuristic, populating `evidence_pro[]`/`evidence_against[]` from `source[]` fields) - When `ambiguity_feed` is empty or all scores are <3, this skill returns `{questions: [], error: null, fallback: false}` — not an error, a legitimate empty result - Re-run is NOT expected: one call per bootstrap session (Q&A is one-shot per Stage 4 invocation) --- ## Input Schema ```yaml ambiguity_feed: Array of insight objects, each: topic: string — short label for the ambiguity (e.g. "project_type") ambiguity_score: number (1–10) — raw confidence gap; higher = more ambiguous evidence_pro: Array<{ source: string, — file:path:line or descriptor snippet: string, — relevant excerpt from the source weight: number — 0.0–1.0 relative weight of this piece of evidence }> evidence_against: Array<{ source: string, snippet: string, weight: number }> ``` **Input constraint:** insights are provided as-is from the synthesis stage; the skill does NOT validate `ambiguity_score` range beyond the ≥3 threshold ...