[Spec 1286] Post-merge state capture: thread narrative + porch completion record

[waleedkadous]

Post-merge state capture for Spec 1286 (follow-up to #1341, which is merged).

Two files, four commits, no code:

File	What
codev/projects/1286-consult-configurable-per-lane-/status.yaml	records the PR-merge and the verify phase transition — the canonical completion record on main
codev/state/aspir-1286_thread.md	the merge + verify narrative, which the thread-retention rule wants on main alongside codev/reviews/

These landed on the builder branch after #1341 merged, so they were stranded there. Opening this
at the architect's direction rather than leaving them behind.

What the verify notes record

The verify phase was a real pass, exercised against the freshly built merged dist/ rather than the
branch:

• Ask 1 (per-lane models): claude → claude-opus-5, codex → gpt-5.6-sol, each returned with
  its provenance key
• Ask 2 (lane selection): PIR → gemini, codex · SPIR → gemini, codex, claude
• Fail-fast: a typo'd lane gives
  Invalid consultation model "codexx" in porch.consultation.byProtocol.pir.models…
• --model-id is live in the merged CLI's --help — the check that mattered most, since that
  flag originally shipped registered, parsed, documented and inert, which is part of why Spec
  1286 exists
• Full suite on the merged state: 4271 passed, 48 skipped, 0 failed; tsc and build clean

Note

The verify-approval gate is still open with Waleed. If it is approved while this PR is open, porch
will add one more status.yaml commit to this branch and it will join this PR — which is the
intended completion record, not drift.

No code, no behavior change, no test impact.

[waleedkadous]

🔄 Closing and immediately reopening this PR to re-trigger CI — no action needed, nothing is wrong with the branch.

The GitHub Actions major outage on 2026-08-06 left workflow run 31121112711 orphaned: status: queued, conclusion: null, zero jobs ever allocated, and no update for ~6 hours. It is unrecoverable — GitHub refuses both gh run cancel ("cannot cancel a workflow run that is completed") and gh run rerun ("this workflow is already running") for the same run object.

Since test.yml and e2e.yml both trigger on pull_request, a close/reopen fires fresh runs without pushing a junk commit, keeping this 8-commit docs PR clean.

The branch is known good: the sibling run 31121112665 completed successfully on this exact SHA (both CLI Tests checks green in ~1 min once it got runners). This is purely an infrastructure retrigger.