v0.2.0

Changelog

All notable changes to agent-router are documented here. The format follows
Keep a Changelog, and the project adheres to
Semantic Versioning.

0.2.0 - 2026-06-15

Added

• 4-level feedback rating (feedback.js): bad/ok/good/excellent → reward
  0/0.34/0.67/1. Interactive keypress rater plus a scriptable direct mode that
  accepts 1-4, b/o/g/e, or words.
• Learning loop (learn.js): discounted Thompson-sampling bandit trains on the
  graded ratings and writes a personal reputation overlay.
• Retrieval index (build-index.js): BM25 + optional dense embeddings, RRF-fused,
  over every installed skill/agent including the full plugins tree. Routing retrieves
  top-K candidates from it instead of eyeballing the session list.
• Auto-update check (update-check.js): daily-cached version check against GitHub.
• Auto web/marketplace fallback: routing runs /skill-finder automatically when
  no installed tool fits the task.
• Shareable/private separation: the repo ships only the curated
  data/registry.json; all personal data (decision log, learned overlay, retrieval
  index) lives under ~/.claude/agent-router/ and is never committed or published.

Fixed

• install.js usage text now says npx claude-agent-router (was npx agent-router).
• The npm tarball no longer includes the machine-specific generated index.
• Plugin manifest: removed the invalid agents field that failed install validation
  (agents: Invalid input). The agent auto-discovers from the agents/ directory.

0.1.0 - 2026-06-07

Added

• Initial release: the /agent-router routing skill, the advisor agent, the
  /skill-finder discovery skill, the curated reputation registry, the npx
  installer, the DAG workflow, and decision logging.