[Task]: Human review and direction for What Next? MCP Epic #282 (Design #284 + Epic close)

Human action required

Review progress and provide direction/approval to advance or close the What Next? MCP Epic (#282):

• Research children #283 and #286 are closed with committed artifacts (what-next-mcp-*.md) and usage reports.
• v1 static decision tree enhancements landed (more state from health/epics/release/curiosity/Goals/fragments, full babysit + base-sync, info-audit #218, Q&A during flows, role hints, cross-ceremony coverage; tests + guidance updated; dogfood in autonomous sessions).
• Fragment committed for the enhancement.
• Detailed "Design #284 input" + draft Epic summary comment + close plan posted in #282 comments (2026-06-07).
• Open: Design #284 (need:refinement per planning), possible remaining Feature polish, then Epic-close ceremony (summary comment on #282 + PR from epic/what-next-mcp to release, human review per AGENTS before starting next Epic).

Why the agent cannot safely proceed

Per AGENTS.md (Human checkpoints, Epic-close ceremony, "stop and let the human review before starting the next epic", "Do not start a new epic autonomously without instruction"): the owner must review the research + design input + draft summary, confirm/iterate the schema/decision tree/ integration points for #284, then authorize the Epic-close. Autonomous agent has already extracted all possible research + v1 code + dogfood + artifacts without inventing owner vision or finalizing the Epic. Creating further Design changes or the close PR would bypass the required human checkpoint.

Context and affected artifacts

• Epic #282 [Epic]: What Next? MCP (Beta Roadmap / milestone #1, area:agent, area:product, risk:medium, Epic: what-next-mcp label)
• Related: #284 (Design), #285 (Feature v1, likely closed), #283/#286 (Research, closed), parent planning in #281 discussion, local docs/research/what-next-mcp-.md + .agentic/releases/unreleased/282-.json + mcp_server.py / agent_guidance.py / tests enhancements + AGENTS mentions.
• Sibling work: Epic #349 / #358-#360 (Task type, which enabled this proper blocker notation), Epic #350 (Test Coverage, one child Research PR #367 just opened+auto-merged).
• This Task itself created during scheduled autonomous loop execution (task 019e9ff27741) while making continuous progress on open issues.

Best-effort instructions

1. Read the latest comments on #282 (esp. the long autonomous progress update with Design input + draft Epic summary + "ready for your review").
2. Review the two research docs and the code changes for plate_what_next (static tree expansions).
3. If the Design input is sufficient, either:
   • Confirm and let autonomous continue (e.g. "proceed with Design doc updates per the input" or "close #284").
   • Or open the Design PR yourself / provide refinements.
4. When ready for Epic close: post (adapted) summary comment on #282 (include outcomes for children, evidence, migration, usage), then perform Epic-close PR per AGENTS §Epic-close ceremony.
5. After merge, the next beta Epics (Test Coverage #350, Human Action Items #349, Release Ceremony #306, etc.) can be sequenced.

Done signal

• Human posts the Epic summary comment on #282 (with optional but welcome) or explicitly comments on this Task "Done — direction provided / Epic close authorized. ".
• Or the Epic-close PR for #282 is opened with reference to this Task.
• If repo truth changed as side-effect of the human work, a follow-up Documentation/Feature PR is still expected.

Related links

• #282 (primary)
• #284, #358 (Task type definition), #349 (Human Action Items Epic)
• docs/design/what-next-mcp-.md (no, the research in docs/research/what-next-mcp-.md)
• .agentic/releases/unreleased/282-what-next-mcp-research-and-enhanced-v1-static.json
• AGENTS.md §Epic-close, §Human checkpoints, §Autonomous Mode

Read the latest comments on #282, especially the Design #284 input and draft Epic summary. Review docs/research/what-next-mcp-*.md, the unreleased JSON artifact, and the named mcp_server.py, agent_guidance.py, and test changes. Done means the human provides direction or authorizes closure, followed by the required Epic-close ceremony.