SystemIdentifierSchema/EventNameSchema admit the empty-segment dot shapes the item-name grammar refuses (a., a..b, trailing dots)

[os-trump]

Observation from the #12194 implementation (stage 1 of #12176) — filed unassigned as a finding, not a defect claim.

What was observed

packages/spec/src/shared/identifiers.zod.ts now carries two dotted grammars side by side:

• METADATA_ITEM_NAME_PATTERN (D1: declare the metadata item-name grammar in spec and refuse it loudly at the publish door — dotted qualified form becomes enforced (#12176 stage 1) #12194, enforced at the metadata publish door): segment-anchored — one or more [a-z][a-z0-9_]* segments joined by single dots. Structurally refuses a., .a, a..b, a bare dot, and every empty segment.
• SystemIdentifierSchema (pre-existing): a character-class grammar — lowercase string over [a-z0-9_.] starting with a letter. It ACCEPTS a., a..b, a.b. and any run of dots after the first character. EventNameSchema in the same file is the same family with the same property.

So the junk-shape family the maintainer-ruled item-name grammar refuses (empty segments, trailing/double dots) remains declared legal for every surface SystemIdentifierSchema's own docblock claims: object names, field names, role names, permission set names, action/trigger names, event keys, app IDs, menu/page IDs, select option values, workflow names, webhook names.

Why it may matter (and why it is only an observation)

The #12176 census's lesson generalizes: a permissive identifier grammar at a declaration surface is where junk identity shapes enter and fossilize. The mechanical fix shape now exists in the same file — derive a segment-anchored pattern from the shared ITEM_NAME_SEGMENT source — so a narrowing would not mint a second grammar.

But narrowing SystemIdentifierSchema is an accept-set change across many surfaces at once, and the #12194 premise discipline applies with more force here, not less: each consuming surface needs its own corpus census (does any legitimate existing event key or app id carry a trailing/double dot?) before any regex moves, and the maintainer has ruled nothing about these surfaces. Filed for triage to size and stage — possibly a late stage of the #12176 chain, possibly not worth doing.

Related: #12176 · #12194 · PR #12244 (where the segment-anchored family landed).

[os-trump]

Triage: route domain:spec, kept finding — the card is right that this is an observation, not yet a defect claim: narrowing SystemIdentifierSchema/EventNameSchema is an accept-set change across many declared surfaces at once with no ruling behind it, and each surface needs its own corpus census first. Grading disposition when it comes up: per-surface census first (any legitimate existing key with trailing/double dots?), then either stage into the #12176 chain (whose grammar source ITEM_NAME_SEGMENT is the ready-made mechanical shape) or record a reasoned not-planned; a clean census making every surface junk-free would put the narrowing on the Clause-② contract-review path, maintainer-ruled. Triage seat session session_01TMpU2R7qfDGR8rHKBay9uJ.

---

Generated by Claude Code

[os-trump]

Triage: graded Task, state pm:blocked (keeping domain:spec).

Blocked-by: #12176

Rationale: narrowing SystemIdentifierSchema/EventNameSchema is an accept-set change across every surface their docblocks claim, and the card's own discipline is right — each consuming surface needs a corpus census before any regex moves, and no maintainer ruling covers these surfaces. That makes this a late stage of the #12176 chain, not an independent dispatch: when #12176 closes, the unlock scan returns this card, and the returning grade decides with the chain's census evidence in hand whether the narrowing is worth proposing (at which point it goes to the decision box as a contract change, Clause-② yes) or whether the segment-anchored grammar stays scoped to the publish door. The mechanical fix shape (derive from the shared ITEM_NAME_SEGMENT source) is recorded and needs no re-derivation.

---

Generated by Claude Code