FE-1573: Construct and explain one real net region from a genuine conversation

[lunelson]

🌟 What is the purpose of this PR?

Cut Mission 7 Step A authority: establish whether Brunch can elicit a genuinely complex operational account of Vestera's multi-line production eligibility and changeovers, conserve that meaning, construct a meaningful Petrinaut region from it, and explain every ordinary behaviour-affecting element through declared basis and recorded effects.

This PR is the authority cut only. It converts the accepted planning record into live MISSION.md, leaves Step B in a non-authoritative packet, and restacks the branch onto Mission 6b for review. Implementation, paid evidence, and the product demo have not begun.

What it establishes: the Step A contract, execution graph, inventory/explanation standard, probe tables, and an opened foundation over Mission 6b's accepted narrowed claim. What it does not claim: a constructed region, a working why operation, inherited Mission 7 evidence, Step B authorization, or any release.

🔗 Related links

• Linear: FE-1573 (internal)
• Stakeholder outcome: FE-1478 (internal)
• Parent: FE-1580
• Mission authority: libs/@hashintel/brunch-agent/MISSION.md
• Future spine: MISSION.next.md
• Step B packet (not execution authority): docs/mission-drafts/7-explainable-construction.md
• Foundation verification retained on the parent: docs/evidence/implementations/voice-resumable-reconciliation/verification.md

🚫 Blocked by

• The parent Voice-reconciliation PR, #9564
• Mission 6b's narrowed foundation was accepted by Lu on 2026-09-07; its three deferred claims remain explicit constraints
• Separate owner acceptance before any Step B amendment

🔍 What does this change?

• Cuts live Mission 7 Step A into root MISSION.md and archives Mission 6.
• Moves remaining closure/demo work into the Step B packet; that packet is not execution authority.
• Updates the future spine so FE-1573 is the tracker projection for construction and explanation, advancing FE-1478.
• Records the restack onto accepted Mission 6b and opens shared implementation and paid Step A runs under the existing budget and proof constraints.
• Changes no product code, tests, or publishable packages.

🏗️ Agent notes

Mission authority is libs/@hashintel/brunch-agent/MISSION.md (live Step A, cut 2026-09-07). The six sections, condensed:

Imperative. Prove the current core elicitation + plugin sdcpn-modelling guidance can elicit, conserve, construct, and explain one accepted Vestera region. Visible goal: talk about multi-line eligibility and changeovers, watch the region take shape, ask why by name, and see workpiece passage, evidential standing, and recorded construction steps. Hand-edited and basis-less controls must refuse honestly. Step A ends at an owner gate, not a release.

Throughline. Genuine conversation → mounted /agents/chat/:instanceId → production ChatAgent → core update_workpiece settles Markdown/revisionId/sha256 → plugin mutation cites settled revision and declared basis → actual browser validates and records independently observed effects → authorized why resolves record → basis → passage → evidence. Parallel A1 carrier, A2 revisions, A3 browser effects, and early A4 history pins join into model-facing why and the genuine tracer; Lu's owner gate precedes any B1/B2/B3 work.

Proof. This PR proves only that the contract is cut and restacked. The authorized Step A proof (adversarial tracer, four probes, two measurements, inventory/utility rubric, behavioral discriminator) has not started. Prospective oracles live in MISSION.md; none are claimed as existing tests.

Constraints. Mission 6b's accepted foundation permits shared implementation and paid Step A calls, but Mission 7 must preserve causal per-step client results, active-submission Stop, and the explicit limits on hydration attribution, post-settlement local withholding, and latency. Preserve Voice marker/lifecycle/cancellation contracts in A2/A3. Mixed update_workpiece + terminating construction batch is forbidden. Petrinaut owns schemas and mutations; no second log, capture ledger, ontology, or projection engine. Budget: US$100, Sonnet-class, conservative 200-call cap.

Fog-line. Compaction survival, genuine reopen/relocation, passage locators, carrier behavior per admitted class, browser pre/post observation, and whether current guidance can conserve Vestera complexity at an affordable cadence. These are probe questions, not permission to shrink the region or skip an owner gate.

Stop or reorient. Stop if a join drops Mission 6b's causal-result, marker, lifecycle, attribution or cancellation constraints or claims its deferred properties as inherited proof; if the tracer needs an answer key or a simpler substitute region; if mixed-batch or uncitable revisions are required; if effects cannot be independently checked; or if work broadens into Mission 9/10/11 scope.

Deferred. Step B packet owns demo/closure. Missions 9/10 own repeat/change/revision breadth. Mission 6b's direct-user hydration attribution, post-settlement durable withholding, and comparative latency remain explicit deferred limitations rather than Mission 7 obligations or inherited passes.

Pre-Merge Checklist 🚀

🚢 Has this modified a publishable library?

This PR:

• does not modify any publishable blocks or libraries, or modifications do not need publishing

📜 Does this require a change to the docs?

The changes in this PR:

• require changes to docs which are made as part of this PR (live MISSION.md, MISSION.next.md, Step B packet, and the restack evidence note)

🕸️ Does this require a change to the Turbo Graph?

The changes in this PR:

• do not affect the execution graph

⚠️ Known issues

• No implementation yet. Reviewers should treat this as a contract review, not a product change.
• Mission 6b is accepted narrowly. Its owner witness passes the local microphone/mutation/resume/active-Stop path while explicitly deferring direct-user hydration attribution, post-settlement durable withholding, and comparative latency. Mission 7 must not claim those properties or reuse the prepared-fixture witness as its own proof.
• Step B is not authorized. The demo script, revision list/diff, and full-region release gate live only in the amendment packet.

🐾 Next steps

• Review and accept or amend this Step A contract.
• Re-pin the complete prompt/tool baseline from accepted #9564, including causal per-step client results, before instrument freeze or paid runs.
• Shared implementation and paid Step A work may proceed within the accepted authority and budget; Mission 7 still owes every named oracle.
• Step B requires a separate owner acceptance and an authority-only amendment commit.

🛡 What tests cover this?

None added. This PR changes mission documents only. Existing parent suites are regression priors, not Step A evidence.

❓ How to test this?

1. Read MISSION.md Status, Imperative, Throughline, Proof, Constraints, Fog-line, and Stop or reorient.
2. Confirm the Step B packet is marked non-authoritative and is not treated as a second live mission.
3. Confirm MISSION.next.md points FE-1573 / FE-1478 at this cut and does not authorize implementation.
4. Confirm no product, test, or changeset files are in the unique diff against #9564.

📹 Demo

None. The product demo is a Step B obligation after the owner gate.

[vercel]

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

Project	Deployment	Actions	Updated
hash	Ready	Preview	Sep 7, 2026 6:13pm UTC
petrinaut	Ready	Preview	Sep 7, 2026 6:13pm UTC
petrinaut-docs	Ready	Preview	Sep 7, 2026 6:13pm UTC

1 Skipped Deployment

Project	Deployment	Actions	Updated
hashdotdesign-tokens	Ignored	Preview	Sep 7, 2026 6:13pm UTC

[lunelson]

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

• FE-1573: Construct and explain one real net region from a genuine conversation #9562 👈 (View in Graphite)
• FE-1580: Reconcile Voice turn behavior on the shared Brunch conversation #9564
• FE-1575: Resume one Brunch workpiece and Petrinaut document across tabs #9537
• FE-1574: Let Voice speak through canonical Brunch conversations #9528 : 1 other dependent PR (#9531 )
• FE-1563: Establish Brunch elicitation skill boundaries #9468
• FE-1525: Establish the runbook-to-workpiece control and expose the construction gap #9405
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.