fix(research): preserve markdown report output

[iamjr15]

Why

Research synthesis required the model to emit hidden provenance bookkeeping before the user-visible report. When provider output was constrained, that bookkeeping could consume the response and leave the report truncated or absent. It also made the generated document pipeline harder to reason about.

What changed

• make the model-authored synthesis output only the canonical GitHub-flavored Markdown report
• derive and validate the durable claim map deterministically from already validated research passes
• derive internal pass summaries instead of asking the model to spend output on them
• keep the same canonical Markdown as the source rendered into the PDF artifact
• document the single-source report and provenance boundary

No database schema, migration, environment, vendor, or deployment-topology changes.

Verification

• pnpm lint
• pnpm typecheck
• pnpm turbo build --force
• pnpm deadcode
• pnpm architecture:check
• pnpm turbo skills:build

Production browser acceptance will be run against the merged deployment: natural-language research request, complete chat report, PDF download and text parity, / deliverable picker, console, and Worker logs.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
cheatcode	Skipped		Aug 5, 2026 6:18am