feat(loops): add research loop skeleton

[ncolesummers]

Summary

• add a durable, fixture-backed research-loop with planning -> researching -> authoring -> done
• route spike + agent-ready webhook deliveries to research-specific simulated, durable, and disabled no-op outcomes
• expose the research skeleton through the neutral root orchestrator, portal fixtures, persisted run projection, Storybook, telemetry, and durable documentation

Closes #43.

Why

Loopworks had a development-specific durable path but no second loop shape proving that the orchestration, webhook, persistence, and portal contracts were genuinely loop-neutral. This slice establishes the research actor and artifact boundaries without pulling forward the Eve subagents, live fan-out, provider evals, or versioned research payload schemas owned by #44-#46.

Contract

Stage	Actor	Timeline	Artifact
planning	research-planner	planning	Research plan
researching	researcher	research	One findings placeholder per subquestion from isolated child sessions
authoring	research-author	authoring	Research document
done	loopworks	done	Completion summary

The implementation persists four run steps and four artifact placeholders using the existing schema. Delivery idempotency is scoped to research-loop; disabled research triggers persist research_loop_noop and never create development no-op events. Raw issue bodies and artifact content are excluded from telemetry.

Review focus

1. Check the transactional idempotency lock in src/lib/loops/research-run.ts, especially concurrent delivery behavior.
2. Confirm the manifest and webhook action sets cannot drift and that development responses remain unchanged.
3. Verify the neutral root fails closed for research stages until Research planner subagent #44-Research author subagent #46 declare their actors.
4. Inspect persisted-to-portal projection of researching and authoring; the fixture and database paths should render identically.

Validation

• bun run validate
• bun run build
• bunx eve info
• focused research-loop, manifest, observability, webhook, PGlite, routing, portal, and run-record tests
• two independent adversarial reviews; concurrency idempotency, manifest/action drift, and persisted timeline projection findings were fixed
• agent-browser validation at 390x844, 1280x832, and 1440x960 for the main and disabled-loop paths
• Playwright axe scans in light and dark

Final validation: 27 Playwright tests passed; formatting, linting, agent docs, Markdown, TypeScript, Vitest, and Storybook gates passed. Storybook emitted its existing bundle-size advisory only.

Browser evidence

Research run

Mobile	Laptop	Desktop

Disabled research routing

Mobile	Laptop	Desktop

Scope boundaries

• no database migration
• no research Eve subagents or live child-session fan-out
• no provider evaluations or versioned loopworks.research_*.v1 payloads
• no approval, agent-plan, GitHub writeback, or reusable-component redesign

Co-authored-by: OpenAI Codex GPT-5 noreply@openai.com

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
loopworks	Ready	Preview, Comment	Jul 22, 2026 4:32am

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.