Skip to content

Agentic Design System v1.2.0

Choose a tag to compare

View all tags
@aa-on-ai aa-on-ai released this 17 Jul 20:31
· 83 commits to main since this release
v1.2.0
0eeaa40
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Agentic Design System v1.2.0

This release makes the reasoning behind consequential UI decisions inspectable without putting
another model or browser pass inside the build loop.

  • Captures exact skill and source hashes once before substantial builds, distinguishing files
    observed in execution from files that were only declared.
  • Verifies decision excerpts, product constraints, artifact locations, evidence links, and review
    status once after the final artifact stabilizes.
  • Rejects invented excerpts and prevents declared-only context from masquerading as verified
    causality.
  • Adds an idempotent decision-provenance block to run reports and a worked loop-demo pilot.
  • Keeps routine work fast: tiny fixes skip provenance, substantial runs trace 3–7 decisions, and
    each deterministic operation has a 250ms budget with zero added model, browser, or network calls.
  • Extends clean-install and release checks across Claude Code, Codex, Cursor, OpenClaw, and Hermes.

The preserved loop demo predates this feature, so its pilot rows remain reviewed. Prospective runs
can produce verified receipts from a true pre-build manifest.

See CHANGELOG.md for the complete change inventory.

Assets 2
Loading