Second Brain

Adopt Google Labs canonical DESIGN.md (8-section + YAML frontmatter, `@google/design.md` v0.1.1 alpha) as the single source-of-truth design-token format for the Stitch agent-automation path; retire ou

2 min read

Adopt Google Labs canonical DESIGN.md (8-section + YAML frontmatter, @google/design.md v0.1.1 alpha) as the single source-of-truth design-token format for the Stitch agent-automation path; retire ou

Decision

Adopt Google Labs canonical DESIGN.md (8-section + YAML frontmatter, @google/design.md v0.1.1 alpha) as the single source-of-truth design-token format for the Stitch agent-automation path; retire our custom 7-section format and 9-to-7 mapping; standardize on npx -y @google/design.md@latest lint|diff|export|spec as the pre-create_design_system validation+export gate; keep design-library 9-section open-design exemplars as input (no rewrite of 111 brand files) and adapt them via new 9-to-8 mapping at the single point of consumption (Stitch skill). References canonical example externalized to references/canonical-design-md-example.md per setup-curator progressive disclosure.

Rationale

See synthesis thought a45770e1 — three irreducible requirements (machine-validated lint replaces agent-diligence; single canonical naming; zero migration cost on 111 exemplars), adversarial check rejected the build-our-own-validator alternative on Law-7/Law-10 grounds, premortem alpha-spec-evolution risk mitigated by npx @latest (no version pinning), reversibility = pure git revert. Live verification before AND after refactor (spec fetch + lint smoke + pristine grep + mirror diff -r) all passed. File size 610 vs 600 hard cap is acceptable per setup-curator 1M-context guidance (signal density, not line count, triggers refactor).

Alternatives Rejected

Outcome

Pending

Graph View

Backlinks