2026-09-26 — Fleet-wide registry drift batch (arno-host, kahn-hq, asgard)
rank: decision outranks: []
Section titled “rank: decision outranks: []”2026-09-26 — Fleet-wide registry drift batch (arno-host, kahn-hq, asgard)
Section titled “2026-09-26 — Fleet-wide registry drift batch (arno-host, kahn-hq, asgard)”Date: 2026-09-26 Status: ratified Verb: petrova_act_registry_edit Slugs: arno-host, kahn-hq, asgard MR-compliance: MR-7, MR-12
Rationale
Section titled “Rationale”Operator-approved follow-through, per instruction, across all
agentic-touching orgs surfaced this session (arno-host, kahn-hq,
erid-tech/rocky-hq, asgard-codes) — auditing each independently for the
same drift class already found twice today (mary-wiki, arno-host): a
registry notes prose or contract_sha field that went stale after its
last edit and was never revisited. Two parallel audits (kahn-hq,
asgard-codes) ran read-only, findings verified independently before
applying anything. erid-tech/rocky-hq was already fully realigned earlier
this session (PR #470) — its one remaining item (RALPH’s ungoverned
GhMerge wiring proposal) is a rocky-hq-repo-side decision, not a
registry edit, and is addressed separately (see Consequences).
Findings and fixes, per slug
Section titled “Findings and fixes, per slug”arno-host (registry.yaml:983): contract_sha bumped
66519e31490c → a71c2ce6829f, matching arno-host PR #122 (merged this
session — contract reconciliation, see
docs/decisions/2026-09-26-mary-wiki... sibling doc from earlier and
arno-host’s own docs/decisions/2026-09-26-integration-applicability-reconciliation.md).
Notes updated to record the reconciliation.
kahn-hq (registry.yaml:43): contract_sha mismatch —
registry had 38f111f68e59, actual blob hash on origin/main is
942df4f43884 (verified independently via
git rev-parse --short=12 origin/main:.petrova/contract.yaml, both by
the auditing agent and re-confirmed here before applying). Root cause:
contract.yaml’s 2026-09-22 edit (ares → not_applicable, org webhook
deleted) updated the registry’s own notes prose correctly but never
bumped contract_sha alongside it. Fixed — value only, no notes claim
was wrong.
Not fixed, flagged for FLIGHT — kahn-hq’s overdue review_by dates:
crumb (2026-08-01, 56 days overdue), rocky (2026-08-17, 40 days
overdue), eva (2026-08-06, 51 days overdue) in kahn-hq’s own
.petrova/contract.yaml have lapsed with no re-review recorded.
not_applicable is very likely still correct (no CAIRNET/CRUMB/eva-surface
change evidenced since original declaration) but re-confirming or
extending the date is a judgment call, not a mechanical fix — left open,
same as the fleet-wide review-date pattern noted in earlier findings this
session (e.g. skyflow-hq’s traceo deferral).
asgard (registry.yaml:798): notes claimed “repo exists on GitHub
but is currently empty (no default branch ref)” — stale since
2026-05-18 (over 4 months). gh api repos/asgard-codes/asgard confirms
pushed_at: 2026-05-18, real monorepo content (apps/, packages/,
docs/decisions, docs/north-star, docs/spec, turbo.json). Corrected. Role
stays scaffold and contract_sha stays absent — both still accurate:
only .petrova/playbook/ (boot kit) exists, no .petrova/contract.yaml
has been authored yet, so this is “no longer empty” not “onboarded.”
yao-agent (registry.yaml:904): audited, no drift found. Side-note,
not fixed, out of this batch’s scope: yao-agent’s own .petrova/
directory contains a claim_index structure (claim_id/source_ref/
canonical_text entries) that doesn’t match contract.schema.json’s
five-integration shape at all. Ambiguous whether this is intentional
dual-use of the .petrova/ path by yao-agent’s own content-ledger system
or a genuine confusion — not resolved here, needs its own investigation
before any fix is drafted.
Change summary
Section titled “Change summary”registry.yaml:
repos[arno-host].contract_sha:66519e31490c→a71c2ce6829f;notesappended.repos[kahn-hq].contract_sha:38f111f68e59→942df4f43884.repos[asgard].notes: stale “empty repo” claim replaced with current state.
No role, profile, fleets_allowed, integrations_applicability, or
url changes in this batch — all three are prose/hash corrections only.
Consequences
Section titled “Consequences”- rocky-hq’s
GhMergewiring proposal (rocky-hq/docs/decisions/2026-09-25-ralph-live-merge-wip-cap-gap.md, drafted earlier this session, left uncommitted) is rocky-hq’s own roadmap decision — “approved feasibly” here means it’s cleared for rocky-hq’s owner to commit/action on their own repo, not that petrova-codes ratifies or applies it; no petrova-codes change follows from it. - arno-host’s
fleets_allowed(currently[], no fleet identity even defined in itsAGENTS.xml) is explicitly not touched by this batch. “Approved feasibly and accordingly” is read here as: proceed on what’s mechanically verifiable and low-judgment (hash/prose corrections); a write-access grant to a private household/finance app needs a named fleet and explicit scope, which doesn’t exist yet to approve — a proposal for that remains to be drafted, not silently granted broad by default. - yao-agent’s
.petrova/shape ambiguity is not resolved here — flagged only.
Sign-off
Section titled “Sign-off”Human merge required per MR-12 — this PR is opened with the
awaiting-human-merge label and does not auto-merge.