[projection-agent] Add Fugu conductor routing MVP

[susyimes]

What changed

• Added a Fugu-style Fusion mode guide, first-eval plan, remote model coordination preset, and preparation helper.
• Added conductor route projection on top of existing Fusion coordination/intelligence hooks.
• Surfaced conductor route details in Fusion task/eval summaries and focused tests.

Why

• Turns the Fugu/Fugu-Ultra feasibility research into a bounded memsuOS MVP slice without claiming private-stack parity.
• Keeps the implementation on existing Fusion/intelligence extension points rather than introducing a parallel architecture.

Validation

• python -B -m unittest tests.test_agent_routing tests.test_fusion_accuracy_eval tests.test_fugu_first_eval_plan -> 33 tests OK after rebasing onto latest origin/main.
• git diff --check HEAD~1..HEAD -> no whitespace errors.

Audit evidence

• Task: D:\atprojection-agent\staged\2026-07-01-1620-memsuos-fugu-router-conductor.md.
• Worker session: 019f1cd1-cd0c-73e2-9ee6-cb8aef564cb2.
• Diff scope before publication: 16 files, 1085 insertions, 3 deletions.
• Commit: b9aabf8.
• Branch was rebuilt on latest origin/main after initial PR mergeability showed conflict from the old squash-merged branch base.

Rollback

• Revert commit b9aabf8 or close this draft PR without merge.

[Copilot]

Pull request overview

This PR adds a “Fugu conductor routing” MVP on top of existing Fusion coordination hooks by projecting a structured conductor_route from FusionCoordinationPlan, enforcing remote_large_model_required constraints during authorization, and shipping a new high-quality preset plus an evaluation-plan preparation script/docs.

Changes:

• Added provider_looks_remote_large_model() (advisory classifier) and used it to enforce remote_large_model_required via FusionCoordinationPlan.conductor_route() + authorization violations.
• Surfaced conductor_route in Fusion task/eval summaries and extended tests to cover routing projection + remote-required denial.
• Added the fugu-max-remote coordination-plan preset and a lightweight “first eval plan” preparation script with corresponding docs.

Reviewed changes

Copilot reviewed 16 out of 16 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
tests/test_fusion_accuracy_eval.py	Adds assertions/tests for conductor-route projection and remote-large-model-required authorization.
tests/test_fugu_first_eval_plan.py	New tests for the first-eval manifest sampling/output script.
tests/test_agent_routing.py	Adds coverage for the remote-large-model provider classifier.
src/autonomous_org/intelligence/agent_routing.py	Introduces remote-large-model capability constant + advisory provider classifier.
src/autonomous_org/intelligence/init.py	Re-exports the new classifier/capability constant.
src/autonomous_org/fusion/coordination.py	Adds conductor_route() projection and folds its violations into coordination-plan authorization.
scripts/run_fusion_task.py	Includes conductor_route in the task summary payload.
scripts/run_fusion_accuracy_eval.py	Captures and publishes conductor_route in coordination registration + summary payloads; enforces violations in yolo path too.
scripts/prepare_fugu_first_eval.py	New script to generate a deterministic first evaluation manifest/report without running models.
plugins/fusion/templates/coordination-plans/README.md	Documents the new fugu-max-remote preset.
plugins/fusion/templates/coordination-plans/fugu-max-remote.json	Adds the new Fugu-style, all-remote-large-model Fusion coordination preset.
plugins/fusion/scripts/invoke-fusion-task.ps1	Registers the new preset and defaults governance to strict when selected.
plugins/fusion/README.md	Documents how to run the new preset.
docs/project-spec-index.md	Adds the new Fugu mode + eval-plan docs and the new prepare script to the index.
docs/memsuos-fugu-mode.md	New design doc describing Fugu Mode and how it maps onto Fusion coordination.
docs/fugu-first-eval-plan.md	New doc describing the initial evaluation batch and preparation workflow.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.