Skip to content
PETROVA

Proxy countersign is retired — human countersign means a commit authored by the human

Date: 2026-09-05 Status: open — awaiting human countersign Supersedes: docs/decisions/2026-08-15-proxy-countersign.md Superseded-by: none — current

Decision

Section titled “Decision”

Proxy countersign is retired, effective today. docs/decisions/2026-08-10-blanket-countersign.md (mary-wiki) is retired with it — that record was the same instrument at a larger radius, twenty-two decisions signed by one blanket instruction with “no per-decision review” in its own words. Both existed to solve a throughput problem: today’s three T7 closes (2026-09-05-sign-off-heading-synonym.md, 2026-09-05-ratified-status-mapped-to-closed.md, 2026-09-05-accepted-published-resolved-and-proposed-mapped.md) show it doesn’t exist — each was directed per-doc, in session, in seconds. From here, a human countersign is a commit authored by the human. No proxy tick, no blanket instruction, no agent-as-scribe.

The 2026-08-15 ruling that authorised proxy ticks was itself never countersigned — left deliberately unticked, on the correct instinct that “a ruling about the limits of proxy countersign that was itself countersigned by proxy would be the weakest possible instrument.” That unsigned authority has now stood for three weeks and been cited, unexamined, by three more closes today. Retiring it removes an unexercised authority rather than exercising it for the first time to bless itself — cheaper than signing it, and it closes the same self-referential gap the original ruling’s own footnote named and declined to close.

Consequences

Section titled “Consequences”
  • Historical ticks are untouched. The thirteen-plus (proxy) marks made under the 2026-08-15 and 2026-08-10 conventions stay as-is, per MR-7 — reverting a working state to a stale one to make a ruling look clean would be the forgery of sequence this corpus already refuses elsewhere (F4-12). They remain real signatures, correctly marked as proxy, for records that predate this retirement.
  • Enforced, not conventional. countersignDefects() in cli/src/verbs/validate.ts gains a third defect class, proxy-after-retirement: a (proxy) tick on a doc dated 2026-09-05 or later is refused. A hand signature is untouched regardless of date. Falsified red-before-green: the check disabled, the three new positive cases failed; re-enabled, all 23 tests in cli/tests/verbs/validate-countersign.test.ts pass.
  • docs/decisions/2026-08-15-proxy-countersign.md’s Superseded-by field is updated to point here (metadata edit, same footing as a Sign-off tick under MR-7). Its body is not otherwise touched — it accurately describes a convention that was real while it lasted.
  • docs/decisions/2026-08-10-blanket-countersign.md (mary-wiki) needs the same Superseded-by pointer, filed there since it’s that repo’s own MR-7-governed artefact.

Sign-off

Section titled “Sign-off”
  • Subagent: fleet (mary-wiki/petrova-codes session 2026-09-05), authored under the operator’s directive in session (“Kill it,” “supersede the 2026-08-15 decision,” “human countersign means a commit authored by you”)
  • Human: FLIGHT<alex@devarno.com:2026-09-22> — approves proxy countersign’s retirement, the retirement of 2026-08-10-blanket-countersign.md with it, and that historical (proxy) ticks stay unedited