Skip to content
PETROVA

Mímir retired — Q1 resolved by retirement, not superseded/absorbed/coexistent

Date: 2026-09-25 Status: closed Supersedes: none Superseded-by: none — current

Context

Section titled “Context”

2026-09-22-mimir-reconciliation-scoping.md scoped four questions before asgard-codes/mimir was wired into anything or petrova-codes changed its own charter in response. Human ruling this session, after a direct investigation into what Mímir actually is (not what its own docs claim):

Retire it. Per the human’s own framing: “is MÍMIR in itself overkill? Are we better off removing than adding more logic?” — approved after the investigation below.

What the investigation found (verified against code, not docs)

Section titled “What the investigation found (verified against code, not docs)”
  • Charter (asgard-codes/mimir/README.md): 6 governance-of-governance routines (G0-G5 — fleet reconcile, MR-audit, seam-gating, traceability audit, drift/realign, onboarding). Only G0 has code.
  • mimir-reconcile.yml ran hourly, “success,” for the entire span — but its own header comment says DRY-RUN-HONEST: the PETROVA diagnose verb is not bound, so every repo is written UNRESOLVED. A green check that means “looked at nothing,” running since ~2026-07, is a false-confidence generator, not paused work — the same class of failure this org’s own docs/decisions/2026-08-17-the-projection-stops-restating-phase-status.md (arno-host) and the vacuous-CI-gate finding in this repo’s own docs/audit/state.md already caught elsewhere.
  • mimir-core package: last substantive commit 2026-07-27. Two months dormant at time of retirement.
  • mimir/ docs tree: last touched 2026-07-13. Same dormancy window.
  • DEC-ECO-0001/0002 (Mímir’s own founding charter): still PROPOSED, never reached ACCEPTED, in the 2.5 months since Created: 2026-07-08.
  • One real dependent: grace-hq DEC-0009 (2026-08-07, ratified, countersigned) withdrew a grace-hq requirement (M2.PETROVA.1, cross-repo phase coordination) specifically because Mímir “already claims this seam in substance.” That claim was true only in charter text, never in shipped capability.

Decision

Section titled “Decision”

Q1 resolved: Mímir is retired. Not superseded (nothing replaces its role), not absorbed into petrova-codes (its G0-G5 routine set is not becoming petrova-codes capability), not coordinate/consumer (it isn’t kept running as a scoped-down reader). It stops.

This also resolves Q3 (write-access ceiling — moot, nothing is writing) and closes the acute half of Q4 (grace-hq’s dependency risk) by putting grace-hq on notice directly, rather than leaving it to discover the gap on its own. Q2 (MR corpus drift mechanism) was independently closed this session by the core/templates v1.2→v1.3 sync (petrova-codes/templates#7) — orthogonal to Q1 but no longer relevant to Mímir specifically since Mímir no longer exists to drift against.

This does not re-open grace-hq’s Phase 02-E or re-author M2.PETROVA.1. That is grace-hq’s own call (grace-hq PR #204, filed to put DEC-0009’s cancellation reason on notice, not to decide for grace-hq).

Actions taken (this session, 2026-09-25)

Section titled “Actions taken (this session, 2026-09-25)”
  1. asgard-codes/yao-agent PR #240 (not yet merged): disables mimir-reconcile.yml’s hourly schedule (keeps workflow_dispatch for a manual run), appends retirement entries to DEC-ECO-0001 and DEC-ECO-0002 per their own MR-7 append-only convention, adds a retirement notice to mimir/README.md.
  2. grace-hq PR #204 (not yet merged): new decision doc flagging that DEC-0009’s cancellation reason (Mímir claims the seam) no longer holds — puts the gap on notice, doesn’t decide grace-hq’s response.
  3. This doc.

Both cross-org PRs are open, not merged — merging them is a separate, explicit act this doc does not authorize on its own.

What this doc explicitly does NOT do

Section titled “What this doc explicitly does NOT do”
  • Does not merge either linked PR.
  • Does not decide what (if anything) replaces Mímir’s governance-of- governance role. If a future need for that layer resurfaces, it gets designed fresh against what’s actually needed, not resurrected from unratified scaffolding.
  • Does not touch registry.yaml’s asgard-codes/asgard row — no registry change follows from a retirement inside a consumer repo’s own subtree.

Consequences

Section titled “Consequences”

For code: none in petrova-codes. Cross-repo changes are in the two linked PRs, pending merge.

For docs: 2026-09-22-mimir-reconciliation-scoping.md stays open on record (append-only, not edited) — this doc is the resolution, cited by reference, not a silent overwrite.

For in-flight phases: none in petrova-codes.

For invariants: MR-9 (no invented invariants) — the retirement is grounded in verified code/CI state (dormancy dates, DRY-RUN-HONEST comment, PROPOSED status), not assumed from the charter’s own prose.

References

Section titled “References”
  • docs/decisions/2026-09-22-mimir-reconciliation-scoping.md
  • asgard-codes/yao-agent PR #240
  • grace-hq PR #204
  • docs/decisions/2026-09-25-arno-host-final-interfacing-governance-layer.md (orthogonal ruling, same session)
  • ~/code/workspace/asgard-codes/mimir/README.md, DEC-ECO-0001/0002, .github/workflows/mimir-reconcile.yml (read directly, this session)
  • ~/code/workspace/grace-hq/docs/decisions/2026-08-06-02-phase-02e-cancelled-mimir-seam.md (DEC-0009)

Sign-off

Section titled “Sign-off”
  • Subagent: claude-sonnet-5 (session_01488gmoyur1UAMfpNMgUAhC)
  • Human: alex@devarno.com — confirms retirement (not supersession/ absorption/coexistence) is the correct Q1 answer, and that the two linked cross-org PRs should be merged.
  • Countersigned by human:alex@devarno.com on 2026-09-25.