Skip to content

v0.7.0

Choose a tag to compare

View all tags
@kiloloop-release kiloloop-release released this 21 May 02:46
· 2 commits to main since this release
v0.7.0
178148d
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

What's New in v0.7.0

Added

  • Frontend/UI task category with separate content-patch (15/25/40) and page-build (40/60/90) bands.
  • App-development task category with a generic cold L-style prior and app/UI human-comparison multiplier.
  • 3-round review mode with a 35 minute additive review tier.
  • METR threshold entries for Opus 4.7 (current) and GPT-5.5; opus_4_x retained as a forward-compatible alias.
  • Opt-in structured audit logging via AGENT_ESTIMATE_AUDIT_* environment variables, emitting secret-scrubbed JSON events to stdout, stderr, or a file.

Changed

  • Research-grounded brainstorms now route to the research band instead of the flat brainstorm band.
  • Codex model-key alias now resolves to the GPT-5.5 METR threshold; GPT-5.4 remains available.
  • Corrected the Codex skill install path in skills/estimate/README.md to .codex/skills/....
  • Version bumped to v0.7.0 across package, plugin, action, issue template, and tests.
  • Claude runtime /estimate skill refreshed to v0.7.0 parity with the Codex slice (frontend/app_dev types, 3-round review mode, refreshed METR keys).
  • claude / claude_opus model-key aliases now resolve to opus_4_7 (Opus 4.7); opus_4_6 retained for backward compatibility.

This release also brings the repository structure up to date: a Makefile for dev shortcuts, a scripts/preflight.py pre-PR check, and the multi-runtime skills/estimate/ layout (shared spec + per-runtime Claude/Codex slices).

Assets 2
Loading