source-preflightlisted

# source-preflight — source preprocessing acceptance gate (zero semantic LLM) Run the deterministic CLI preprocessing chain on a candidate source and decide whether its staging artifacts are good enough to enter `ingest`. This skill is a thin wrapper: it only orchestrates `scripts/pipeline.py` and does **no LLM semantic splitting / unit planning / page writing**. Project truth: `CLAUDE.md` / `AGENTS.md`. Engineering format: `docs/skill-runtime/skill-standard.md`. ## 1. Triggers / Non-triggers - **Triggers:** "preprocess this PDF first", "run source-preflight", "build the source profile first", "see if it can be ingested", "just run up to workorder". - **Non-triggers:** "add to the KB / index it" with page writing (use `ingest`); "summarize/split this book" without ingesting (use `source-xray`, published content only); "query the wiki" (use `kb-query`); "translate/explain" (a normal answer). ## 2. Inputs - The user gives: file path `<path>`, domain `<domain>`; `<fmt>` inferred from the extension (pdf/md/docx/pptx); `<src>` derived from the filename. Confirm `<src>` and `<domain>` once. - Read: the **zero-LLM preprocessing constraint** in `CLAUDE.md` / `AGENTS.md`, and `docs/skill-runtime/schema.md` (to understand the workorder write boundary). ## 3. Outputs - `pipeline-workspace/staging/<src>/{source.md, blocks.jsonl, chapters.json, parse_report.json, reconciliation.json, windows.jsonl, workorder.yaml, preflight_eval.json}` + hard-page PNGs / figure assets. - An optiona