infra: clear outstanding validate --all violations at their source

[marc0olo]

Problem

npm run validate (scripts/validate.js --all) reports violations on main that no PR ever surfaces, because CI validates only the files changed in a PR. That CI scope is deliberate: it exists so known pre-existing violations do not block unrelated PRs. The fix is therefore to clear the violations at their source, not to change the CI scope.

State at the time of filing: 11 errors.

Count	File	Error	Where the fix belongs
7	docs/references/internet-identity-spec.md	em-dash in prose	Upstream (dfinity/internet-identity)
2	docs/guides/backends/data-persistence.mdx	mo:base is banned	Validator (false positive)
1	docs/guides/backends/data-persistence.mdx	em-dash in prose	Internal
1	docs/guides/digital-assets/chain-key-tokens.mdx	em-dash in prose	Internal

Status

• The 4 hand-written ones, in fix: clear hand-written validate --all violations, correct the mo:base check #357: two em-dashes replaced, plus a fix to the mo:base rule.
• The 7 in the synced II spec (below).

The mo:base rule was inverted

checkForbiddenPatterns skips fenced code, so the old /mo:base/ pattern could only match prose. An actual import Buffer "mo:base/Buffer"; lives inside a fence and was never flagged, while naming the legacy library in prose was. Both hits on main were the base-to-core migration table in data-persistence.mdx, which is legitimate and mirrors the mapping in AGENTS.md. #357 changes the pattern to the import path mo:base/ and checks it inside fences.

The 7 remaining: fix in dfinity/internet-identity

They are not ours to edit. docs/references/internet-identity-spec.md is generated by scripts/sync-ii-spec.mjs from .sources/internetidentity/docs/ii-spec.mdx, which contains exactly the same 7 em-dashes. Editing the output is pointless: the next sync restores them. They arrived with the release-2026-08-21 spec sync (#350), mostly in the new ii-app-metadata section.

Two ways to land it:

1. Propose the change upstream in dfinity/internet-identity, so both repos converge and nothing has to be rewritten on sync. Preferred, but it asks another team to adopt a style rule that is ours.
2. Normalize during the sync, the way scripts/postprocess-motoko.mjs already replaces em-dashes for Motoko. Entirely in our control and needs no coordination, but the published spec text then differs in punctuation from the canonical spec, which is a heavier thing to do to a specification than to a guide.

These are not exclusive; Motoko does normalization and upstream carries most of the style contract. Suggested order: propose upstream first, and add sync normalization only if it is declined or takes too long.

Instructions to update alongside

AGENTS.md should state the policy once it is decided: which synced targets are normalized at sync time, and that a new synced target has to be registered in the validator's SYNCED allowlist (which today lists only docs/languages/motoko, while the II spec output is not listed and so is checked like a hand-written page).

Related

• fix: clear hand-written validate --all violations, correct the mo:base check #357 clears the 4 hand-written violations
• docs: small frontmatter and style contract so docs/ can sync into developer-docs certified-assets#124 asks a third upstream to drop em-dashes for the same reason, so whatever is decided here should be consistent with it
• infra: reduce .sources/ submodules to what the site actually needs, watch releases for the rest #355 covers the wider .sources/ strategy and owns the AGENTS.md changes