Skip to content
v0.3.0

Your curious AI development partner: test-driven quality, independent review, minimal, private, PR-only, and it learns your project every day.

Changelog

All notable changes to Katsuo are documented here. The format follows Keep a Changelog, and this project uses Semantic Versioning.

[0.3.0] - 2026-11-09

Added

  • Deterministic quality guards in hooks/guard_git.py:
    • secret guard — blocks commits and pushes that introduce high-confidence secrets (private keys, GitHub/OpenAI/AWS/Slack/Google tokens);
    • spec guard — blocks an agent from pushing a feat/<slug> branch without docs/specs/<slug>.md.
  • Tests for both guards.

Changed

  • Moved "no spec, no ship" and "no secrets" from prompt-level rules to execution-level hooks that cannot be skipped.

[0.2.0] - 2026-11-09

Added

  • Learn (retrospect) step. After each change Katsuo records durable, project-specific lessons into docs/memory.md and docs/decisions/, and asks what to remember next time.
  • Start from memory. Every run reads project memory before sizing the work, so Katsuo improves every day instead of re-learning the same things.
  • MIT license and this changelog.
  • A test asserting the learning loop and license are present.

Changed

  • Positioning: quality-first, test-driven, minimal, private, and it learns.

[0.1.0] - 2026-10-06

Added

  • /katsuo command: size → setup → spec → test-driven build → independent review → Pull Request.
  • Independent read-only reviewer agent.
  • guard_git.py PR-only Git hook (blocks protected-branch pushes, force/bulk pushes, and gh pr merge).
  • Deterministic and live end-to-end proof suite.