Skip to content

v0.13.0 — second-agent benchmark, extraction fixes, dependency majors

Choose a tag to compare

View all tags
@zk-xyz zk-xyz released this 17 Jul 22:23
v0.13.0
5ee85b6
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Task-completion evidence release.

First live benchmark numbers

  • Second-agent job completion: 5/5 — a fresh model given only brand_context output produced content the real brand_check + compliance gate accepts (0 flags, ~144 output / ~325 context tokens per artifact)
  • Negative routing: 0% false-positive invocation (0/7) on vocabulary-baited prompts
  • Published with caveats in eval/RESULTS.md

Extraction fixes (#35, corpus expectations flipped, no thresholds lowered)

  • Dark chromatic brand colors (navy/forest/burgundy) can now become primary candidates (HSL-based chromaticity)
  • Generic web-safe fallbacks (Arial et al.) no longer leak from font stacks; corpus precision threshold restored to 0.95
  • Instruction-shaped extracted color names replaced with generated names

Dependencies — individually triaged

  • vitest 4, actions/checkout + setup-node v7 (SHA re-pins), puppeteer-core 25, pdfjs-dist 6
  • TypeScript 7 typechecks the project clean; parked for a standalone migration PR (#18)

Full changelog: https://github.com/Brandcode-Studio/brandsystem-mcp/blob/main/CHANGELOG.md

Assets 2
Loading