FE-835: Orchestrator demo seeds from plan fixtures

[kostandinang]

Reverse the three brunch cook plan fixtures (parallel-utils, layered-todo, resilient-pipeline) into completed-spec fixtures seeded into the intent-graph DB, so the orchestrator can be driven end-to-end SPEC → plan (build-architect, FE-829) → cook — not just from a hand-authored plan.yaml.

• Reverses each plan.yaml into a reviewable SPEC.md/PLAN.md + an *-spec seed scenario (npm run seed -- <name>-spec); adds a blank *-kickoff for live-driving the interview.
• closed-spec-builder.ts: content-agnostic four-phase interview replay (grounding → design → requirements → criteria) with a per-turn impact spread.
• Spec graphs are observer-faithful — grounding/design provenance, constraints, and verifying criteria only; no requirement→requirement execution order (plan truth the architect re-derives downstream).
• Adds an ORCHESTRATOR.md runbook.

How to seed them

Each fixture is registered as a completed-spec scenario. Seed one by name, which prints the new project id:

npm run seed -- parallel-utils-spec      # or layered-todo-spec / resilient-pipeline-spec

Use <name>-kickoff (e.g. parallel-utils-kickoff) instead to seed a blank spec and drive the four-phase interview live from scratch.

To drive the full SPEC → plan → cook loop for all three (build-architect re-derives the slice DAG the fixtures hand-author):

for s in parallel-utils-spec layered-todo-spec resilient-pipeline-spec; do
  id=$(npm run seed -- "$s" | grep -oE 'project [0-9]+' | grep -oE '[0-9]+' | tail -1)
  echo "▶ $s → spec $id"
  npx tsx src/server/cli.ts plan "$id"                                    # architect emits plan.yaml
  npx tsx src/server/cli.ts cook "$(pwd)" --spec="$id" --policy=parallel  # add --petrinaut-stream to watch live
done

These are greenfield specs (cook runs in empty per-slice worktrees). See ORCHESTRATOR.md §7 for the full runbook.

Verification

npm run verify (check + test + build); in-memory seed asserts four phases closed, policy-valid edges, and zero requirement→requirement depends_on.

[cursor]

PR Summary

Low Risk
Changes are documentation and dev/demo fixture seeding with no production runtime or auth paths; main risk is fixture graph drift vs. architect behavior, covered by existing verify/seed assertions.

Overview
Reverses the three orchestrator cook demos (parallel-utils, layered-todo, resilient-pipeline) into fixture-level SPEC.md / PLAN.md and DB seed scenarios so demos can run SPEC → brunch plan → brunch cook instead of only hand-authored plan.yaml.

Adds closed-spec-builder.ts to replay a full four-phase, all-closed interview from an arbitrary knowledge graph (per-item grounding/design turns with impact, accepted requirements/criteria reviews, policy-checked edges). cook-fixture-specs.ts maps each reversed fixture into that graph and registers *-spec scenarios (plus blank parallel-utils-kickoff); scenarios.ts exposes them via npm run seed.

Seeded graphs are observer-faithful: provenance, constraints, and verifies only—no requirement→requirement execution depends_on, so the round-trip exercises the build-architect re-deriving fan-out, joins, gates, and halt shapes from content.

ORCHESTRATOR.md documents cook/Petrinaut workflows and the seed → plan → cook loop for all three specs.

^{Reviewed by Cursor Bugbot for commit 7cd7da7. Bugbot is set up for automated code reviews on this repo. Configure here.}

[kostandinang]

• FE-835: Orchestrator demo seeds from plan fixtures #192 👈 (View in Graphite)
• FE-829: Build-architect for spec-derived cook plans #185
• FE-827: Policy-selected orchestrator slice layout + greenfield promotion-back #182 : 1 other dependent PR (#183 )
• FE-826: Orchestrator resolves greenfield/brownfield from the spec-derived plan, not plan location #181
• FE-819: Orchestrator <> Petrinaut integration improvements #176
• FE-815: Orchestrator demo fixtures: parallel, diamond, halt #174
• FE-813: Cook evaluator runs the verification tests instead of an LLM verdict #170
• FE-800: Spec to orchestrator plan emitter — generate plan.yaml from a completed specification #167 : 1 other dependent PR (#168 )
• FE-764: Brunch<> Petrinaut stream — ephemeral orchestrator hosted SSE server #165
• FE-784: Petrinaut export: colour-fold per-slice subnet #160
• FE-763: Petrinaut event stream — initial markings + transition firings #158
• FE-762: Petrinaut-format JSON export of the compiled net #157
• FE-761: Petri-net semantic alignment for Petrinaut visualization #156
• FE-755: Cook codebase mode for brownfield brunch #155
• FE-747: Declarative output routing for branching transitions #154
• FE-745: Epic-scoped sandbox merge for verify-epic #152
• FE-743: Petri parallel execution with resource pools and per-slice sandboxes #149
• FE-738: Petri semantic lanes with two-lane subnet and engine factory #148
• FE-730: Greenfield orchestrator POC with pi-agent dispatch #143
• main

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

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 5799649. Configure here.}

[lunelson]

This looks useful for getting the demo path closer to the real SPEC → plan → cook flow.

One thing I’d keep an eye on: src/server/fixtures/closed-spec-builder.ts is fixture code, but it now knows how to replay a full four-phase interview using many of the same DB primitives as the product flow. That seems fine for seeding, but I’d avoid letting it become a second “interview implementation” that future tests or features quietly depend on. The boundary is clear in the comments today; just calling it out because this helper is now fairly capable.

The base branch was changed.