v0.5.0 — Agent-Friendly CLI

What's New

Agent-friendly CLI overhaul

This release makes the reflect CLI much easier for AI coding agents to use programmatically.

Session & checkpoint IDs everywhere

• reflect sessions now prints session ID prefixes (e.g. [b7f5e89a-ba1]) on every line — agents can chain directly into reflect sessions <id> for detail
• reflect timeline now prints both session IDs and checkpoint IDs (e.g. [cp:af09a953]) — agents can chain into entire explain --checkpoint <id>

--json on most commands

• reflect sessions --json — full session list as JSON array
• reflect sessions <id> --json — session detail as JSON
• reflect search <query> --json — structured search results
• reflect status --json — all status data as structured JSON

New flags

• reflect search --limit N — control max results per source (was hard-coded at 10)
• reflect context --verbose — show subagent progress on stderr

--help with examples

• Every subcommand now includes copy-pasteable Examples: in --help output

Cleanup

• Shared formatters extracted to lib/fmt.py (deduplicated from sessions/timeline)
• Status errors now go to stderr (was stdout)
• Removed unused load_format import from improve.py
• Keeper agent streamlined to reference SKILL.md evidence ladder directly

Full Changelog: v0.4.7...v0.5.0