research

Run the secondary-literature ingest per CLAUDE.md, section "Secondary literature".

For a new file in raw_secondary/:

1. Read it and compile its POSITION (what it claims about the subject of the study) — by PARAPHRASE, with provenance (raw_secondary/<file>) and an anchor phrase for each key claim. Do NOT cite it as a fact about the world, only as someone's position with attribution.
2. Two-layer verification (rule 1): for each attributed position verify not only that the anchor phrase IS in the file, but that the position holds in CONTEXT — that the use does not distort what the author actually claims. Word-match is not enough; distorting a position is fabrication. Uncertain context → weaken or drop.
3. Place it in process/research.md. Map agreements and divergences with positions already there.
4. Mark the gap the study's thesis fills (what this literature overlooks).
5. Recompute originality: what of the thesis does this source already say itself → what remains as the author's own contribution? Write that remainder. (A source that confirms the thesis can quietly take its novelty.)
6. Type and balance: is it a theoretical lens, or a document of field consensus? When a 3rd+ theoretical lens accrues without a single field source, REPORT the imbalance (risk of an "apparatus parade") — what's missing is scholarship about the subject, not another lens.
7. Attribute ONLY what you find in the file — never "the author would probably say". Don't name a source outside raw_secondary/ at all.

Log it. After a few sources, offer to write the positioning for the introduction from research.md.