# Changelog

## [0.2.2] – 2026-06-21

### Changed

- **README** – Removed claim that personality does not affect tool-calling, code quality, or reasoning depth. Evaluation showed the scoping instruction is not fully enforced and personality can influence behavior beyond tone.

## [0.2.1] – 2026-06-21

### Fixed

- **Communication style scoping** – Personality prompts are now wrapped in an explicit framing instruction that restricts them to tone, phrasing, and conversational manner only.

## [0.2.0] – 2026-06-20

### Added

- **Casual personality** – informal, natural, and easygoing communication style
- **None option** – disables personality prompt injection; appears first in the selector

### Changed

- **Teacher personality** – rewritten for model consistency: teaching is now a structural default with a concrete output pattern (1-3 sentences after the task), per-category directives, and a single narrow exception

## [0.1.2] – 2026-06-14

### Fixed

- **Teacher personality** – narrowed teaching skip conditions to simple facts and explicit answers only, with positive guidance on when to teach during implementation work

## [0.1.0] – 2026-06-12

### Added

- **`/personality` command** – TUI selector to switch communication styles at runtime
- **Three builtin personalities** – Pragmatic (from Codex CLI), Friendly (from Codex CLI), Teacher (custom)
- **Custom personality support** – Drop `.md` files with YAML frontmatter into `~/.pi/agent/personalities/`
- **Persistence** – Active personality saved to `~/.pi/agent/personality-state.json` and restored on next session