chore(pm-triage): rubric.md still anchors P0-P3 on the pre-launch flip, which shipped #526

Area: infra · agent tooling — gap · found via the pm-triage routine's own runs

Expected: .claude/skills/pm-triage/references/rubric.md carries the priority anchor the daily triage routine rates every issue against, and instructs "Re-anchor when the milestone moves."

Actual: the file still anchors on a milestone that shipped eight days ago:

Anchor: the imminent public launch — v0.1.0-alpha.1 cut and the repo public-flip, which happen together (tracked in #149).
P0 — blocks the flip. Must be fixed before the repo goes public / the tag is cut.

v0.1.0 was tagged 2026-08-19, the repo is public, and #149 is closed. Nothing can "block the flip" any more, so the written P0 definition selects the empty set.

Impact: the routine has been rating against a post-launch anchor that lives only in a comment on #501 and in the routine's local pending.md —

P0 — live in a shipped, public release and severe enough to cut a patch for. P1 — wanted in the next release. P2 — blocks broader adoption. P3 — backlog.

— i.e. not in the file the next run reads. That working anchor has already priced real issues (#514/#515/#516/#517/#518, and #523/#524 today), and both current P0s (#460, #474) are shipped-release defects, a category the written rubric has no slot for at all. Two consecutive runs recorded that nothing had been mis-rated yet and called that luck; the ratings above are what the luck was covering.

Scope: re-anchor rubric.md's P0–P3 definitions and examples on post-v0.1.0 reality — one file, through the normal gate.

Related: #501 (where the working anchor is recorded), #149 (closed), #522 (last change to a pm-triage reference file)

From the pm-triage all routine, escalated across seven runs; validated by code-read against 7b3c25e on 2026-08-27.