ambiguity-detector

# Ambiguity Detector ## Purpose / When to Activate Activate: - As Stage 3.5 of the `/gaai:bootstrap` pipeline — after `bootstrap-llm-synthesis` returns results and before `smart-question-generator` is called - The bootstrap orchestrator filters `synthesis_result.entries[]` to `clarity: "open-question"` and passes them as `synthesis_entries` - When `synthesis_entries` is empty and `ast_signals` is null, the skill runs heuristic-only analysis on `surface_scan_result` and may still produce entries from language distribution This skill is **pure heuristic** — it makes no LLM calls. It scores ambiguities deterministically from structural signals. All scoring logic executes client-side before the Q&A stage (consistent with any active "LLM stays client-side" and "instructions stay client-side" DECs in your registry). --- ## Input Schema ```yaml surface_scan_result: # required — output of project-surface-scan total_file_count: number languages: Array<{ language: string file_count: number rank: number }> ext_counts: { [ext: string]: number } # e.g. { ".ts": 142, ".vue": 3 } dir_counts: { [dir: string]: number } # depth-1 directory file counts synthesis_entries: # optional — null or empty array acceptable # Array of entries from bootstrap-llm-synthesis filtered to clarity="open-question" # Pass [] or null when synthesis stage was skipped or produced no open-questions - topic: string # e.g. "What framewor...