Verified market report access for AI Agents

quant-market-brief reads an existing public daily report and emits a deterministic quant.market_brief.v1 JSON object. It uses the same content hash, public manifest, source and market-date checks as the public daily-report importer. It does not fetch data, regenerate reports, write files or send messages.

Install the package in your environment, and copy the repository's skills/quant-market-brief folder into your Agent tool's skills directory. The Skill calls the installed CLI; it does not maintain a second data or scoring layer.

quant-market-brief --input /path/to/bundle/daily_report.json \
  --manifest /path/to/bundle/publication.json --expected-date 2026-10-02

Use the original owner bundle, not the reduced public Pages projection. For a news-only revision pass --previous-report /path/to/immutable/daily_report.json. The requested date is the report's market date, not its UTC generation date. “Latest” must be resolved by the caller using a known publication/archival source; the CLI does not search production directories or silently choose an older file.

The output includes:

Field	Meaning
report_date, as_of, generated_at	Requested report date, market cutoff and generation time
source	Bundle report filename and run ID
artifact_hash, content_hash	Exact input file SHA-256 and owner content hash
data_quality, evidence_level, limitations	Preserved quality, descriptive evidence stage and interpretation boundaries
report	Allowlisted public facts, reviewed claims, evidence IDs, sections, source status and missing sources

Exit status is 0 on success; validation/read errors return 2, describe the error on stderr and emit no JSON on stdout. Public claims must reference evidence in the public projection with source URLs. Legacy id-only event projections are rejected rather than treated as complete evidence. The file is read once, so its hash always describes the bytes used for the output. Hash verification establishes consistency with the provided manifest; it is not publisher authentication or independent fact verification. No future-return prediction, counterargument, topic score or watchpoint outcome is manufactured when absent from the report. The five-date Pages window remains unchanged; longer comparisons require explicitly supplied retained bundles.

Legacy scoring meanings

The existing keys, weights, arithmetic and totals remain compatible. breakdown_detail now adds metric, label, evidence_level and limitations. Digest HTML and summaries use the actual metric meaning, including when reading legacy scores without these added fields.

Theme	fundamental	valuation	liquidity	event
AI	Theme performance	Average PE relative to 35	Average PS relative to 8	Fixed 70
Magnificent 7	Aggregate market cap relative to $11.5 trillion	Average PE relative to 32	Average PS relative to 7	Fixed 68
BTC	Fixed 50	Futures basis	ETF net inflow	Fixed 65

PS measures valuation; market cap measures size. Neither measures financial quality or execution liquidity. BTC sentiment uses funding rates, while equity sentiment uses the existing market sentiment aggregate. Missing inputs keep the existing fallback flags and values; fixed constants are explicitly identified. All scores remain descriptive heuristics without out-of-sample predictive validation. This change corrects interpretation, not the underlying scoring model.