← ClaudeAtlas

bootstrap-llm-synthesislisted

Construct the LLM synthesis prompt from project surface scan + optional tree-sitter context + optional Q&A answers. Call the LLM. Parse and validate the response into 6-8 structured memory entries with clarity tags and source traceability. Used as Stage 3 of the /gaai:bootstrap pipeline.
Fr-e-d/GAAI-framework · ★ 147 · AI & Automation · score 82
Install: claude install-skill Fr-e-d/GAAI-framework
# Bootstrap LLM Synthesis ## Purpose / When to Activate Activate: - As Stage 3 of the `/gaai:bootstrap` pipeline (after `project-surface-scan` and optional tree-sitter parse) - When the bootstrap orchestrator is ready to distill raw project signals into structured memory entries - Re-run if Q&A answers (Stage 4) are incorporated after an initial synthesis pass Produces `synthesis_result` — structured memory entries ready for consent gate (Stage 5 pre-write gate) and memory ingest. --- ## Clarity Tag Semantics Each entry carries a `clarity` field. Use these definitions consistently: | Value | Meaning | |---|---| | `confirmed` | Evidence found directly in source files (imports, config, explicit declarations). Source[] must include file:line reference. | | `inferred` | Reasonable conclusion from structural patterns (directory layout, naming conventions, dep graph). Source[] should list supporting files. | | `open-question` | Significant ambiguity remains. Formulate as a question in `content`. Source[] may be empty. | | `declared-by-user` | User stated this explicitly in Q&A answers. Source[] should reference the Q&A input. | --- ## Process ### Step 1 — Assemble synthesis context Build the `synthesis_context` object that will be injected into the prompt: ``` synthesis_context: project_summary: total_files: surface_scan_result.total_file_count size_class: surface_scan_result.size_class estimated_loc: surface_scan_result.estimated_loc top_languages: s