Track B Phase 1 — synthesize market-facing theses from Supabase research only. No portfolio weights, no preferences or investment-profile. Produces market_thesis_exploration JSON. Triggers: portfolio task Phase 1, "explore market theses from today's research".
Turn compiled research into exploratory market theses: opportunities, trends, headwinds/tailwinds, bull/bear sketches, sub-theses, and explicit validation / invalidation criteria. Style is hedge-fund-like idea generation — not yet mandate-constrained and not necessarily actionable as trades.
Do NOT read: config/portfolio.json, config/preferences.md, config/investment-profile.md.
May read (optional universe only): config/watchlist.md — ticker names and categories to keep vehicle ideas realistic; never infer holdings or weights.
daily_snapshots for {{DATE}} — regime, , key levels.segment_biasesdocuments — digest payload; any research-delta/… for the date; research-changelog/{{DATE}}.json if present; research_baseline_manifest if present.baseline_date / week anchor) if published — use as carry-forward base; apply deltas via new full JSON or document_delta targeting this doc (operator choice).The digest is input only here — it must already exist from the research task (cowork/tasks/research-weekly-baseline.md / cowork/tasks/research-daily-delta.md); Track B does not compile it.
Announce: "Market thesis exploration — research-only context loaded."
Write a short overview (body.executive_digest_pointer) that orients the reader: regime, largest movers, what stayed stable. Point to digest paths or segment names — do not paste the entire digest.
Add body.deeper_dives[] strings for 1–3 themes that deserve more narrative (still research-grounded).
For each thesis in body.theses[]:
thesis_id: stable ID for the week (e.g. MT-001) — align with later vehicle map rows.sub_theses — optional nested claims with their own validate/invalidate lines.validation_criteria / invalidation_criteria — measurable, time-bound where possible (prices, spreads, policy events, data prints).headwinds / tailwinds / bull_case / bear_case — exploratory, not a court trial; surface asymmetry.linked_research_refs — cite document_key, changelog lines, or snapshot paths.Emit doc_type: market_thesis_exploration, schema_version: 1.0, date: {{DATE}}.
document_key e.g. market-thesis-exploration/{{DATE}}.jsonpython3 scripts/validate_artifact.py - < payload.jsonpython3 scripts/publish_document.py --payload - --document-key market-thesis-exploration/{{DATE}}.json --title "Thesis Exploration" --doc-type-label "Market Thesis Exploration"linked_research_refs entry.Called from cowork/tasks/portfolio-pm-rebalance.md before skills/thesis-vehicle-map/SKILL.md.