Skip to content

Adopt CUPEL rung-1a (merge trailers), defer rung-1b

Date: 2026-09-25 Status: open Supersedes: none Superseded-by: none — current

stratt-hq/apps/meridian/cupel/BINDING.md (rule 3, “Order of adoption”) defines a two-rung merge-trailer convention. mary-wiki’s spec docs/superpowers/specs/2026-09-25-cupel-rung-1-adoption-design.md (mary-wiki PR #728, approved 2026-09-25) proposes rung-1a adoption across three repos as three separate, file-disjoint PRs — mary-wiki, petrova-codes, panto-cloud — never batched (mary-wiki’s own Draft rules: “ratify one at a time, never batch”). This doc is petrova-codes’ own track (Track B).

Adopt STRATT:CUPEL’s merge-trailer convention (stratt-hq/apps/meridian/cupel/BINDING.md, rule 3) for future ratifying/closing commits in this repo:

Cupel-Red: RUN-<timestamp>-<id>
Cupel-Green: RUN-<timestamp>-<id>
Cupel-Criterion: <criterion-id>

This is rung-1a only (BINDING.md’s “Order of adoption” step 1, “zero tooling, immediate traceability”). Rung-1b (oracle-exists CI check, step 2) is explicitly not adopted here.

Baseline, checked 2026-09-25: git log --all --grep "Cupel-Red:" in this repo returns zero matches. No commit in petrova-codes has ever used this trailer. This decision establishes the convention going forward; it does not retroactively apply to any past commit.

This repo is role: control-plane, fleets_allowed: [] (registry.yaml). This decision is a human-authored convention adopted via normal PR review — it introduces no fleet automation, no new agent, and no new CI tooling (rung-1b, which would add a CI check, is explicitly deferred). It is consistent with the existing registry note that “changes to the control plane go through human PR review only, not agent-fleet automation.”

Why rung-1b is deferred, not adopted alongside

Section titled “Why rung-1b is deferred, not adopted alongside”

oracle-exists needs a machine-parseable criterion → oracle.cmd mapping. Building it before that structure exists risks BINDING.md’s own named failure mode (VP-1): “a check whose exit code nothing consumes manufactures confidence, which is worse than no check at all.” Rung-1b is deferred to its own future decision once a structured oracle-citation format exists to build it against — and, per the Control-plane posture above, any such future CI addition is itself a heavier act in this repo than a plain convention, so it warrants its own review independent of this one.

  • Adopt rungs 1a and 1b together, with a loose parser against whatever criterion text currently exists. Rejected — exactly the VP-1 trap named above: a check built against text not written to be parsed produces false negatives, which is worse than no check, and it would add CI tooling to a control-plane repo before any human review of that specific addition.
  • Skip merge trailers entirely until a structured oracle-citation format exists. Rejected — rung-1a is “zero tooling, immediate traceability” per BINDING.md; there is no structural reason to wait, and it introduces no automation this repo’s control-plane posture would need to review.
  • Fold this into one cross-repo decision covering mary-wiki, petrova-codes, and panto-cloud at once. Rejected — mary-wiki’s own Draft rules require ratifying one at a time, never batching, and the three repos have materially different postures (this repo’s control-plane lockdown; panto-cloud’s standing write-surface refusal) that don’t fit one decision.

Nothing retroactive. Forward-looking: any future rung-1b decision in this repo must name its oracle-citation-format prerequisite as met, and must be reviewed as a distinct CI-tooling change consistent with this repo’s control-plane posture, before proposing the check.

For code: none. No CI, no scripts, no schema changes.

For docs: this doc only. No projection elsewhere in this repo currently lists decision summaries the way mary-wiki’s .claude/rules/decisions.md does, so no companion pointer edit is made.

For in-flight phases: none.

For invariants: none — this is a commit-message convention, not an MR addition, modification, or repeal.

  • stratt-hq/apps/meridian/cupel/BINDING.md (rule 3, “Order of adoption”)
  • mary-wiki docs/superpowers/specs/2026-09-25-cupel-rung-1-adoption-design.md (mary-wiki PR #728)
  • mary-wiki docs/superpowers/plans/2026-09-25-cupel-rung-1-adoption.md (Task 2)
  • registry.yaml (role: control-plane, fleets_allowed: [])
  • Subagent: claude-sonnet-5 (session_01DekDHQxPychKZW5tCUrZ5R)
  • Human countersign — confirms rung-1a adoption (merge trailers) and rung-1b deferral is the correct scope for this repo’s control-plane posture.