v0.4.0 — doctor, score, drift, compare, audit + GitHub Action

The differentiation release. Five new commands turn AgentNotary into a lifecycle health system.

✨ Five new commands

• agentnotary doctor — one-command health scan with actionable punch-list (brew doctor for AI agents)
• agentnotary score [--badge|--json] — single 0–100 governance score across 7 dimensions + shareable shields.io badge URL
• agentnotary drift — re-probe the model, quantify drift since last seal. First OSS tool to detect silent provider weight updates.
• agentnotary compare <a.lock> <b.lock> — high-level lockfile diff with section-by-section state (model, prompts, tools, deps)
• agentnotary audit <session-id> — forensic session security audit (PII detection, repeat-loops, cost spikes, guardrail events)

🎯 GitHub Action

Drop-in CI step:

- uses: CharanBharathula/agentnotary@v0.4.0
  with:
    manifest: agentnotary.yaml
    min-score: "70"
    fail-on-drift: "true"

Runs seal --verify, attack --dry-run, compliance --check, score. Posts a summary to the PR.

🪪 README badge

agentnotary score --badge
# → https://img.shields.io/badge/agentnotary-87/100-brightgreen

Add to your README. Every project that adopts the badge drives discovery back to AgentNotary.

📊 Quality

• 202 tests passing (up from 169 in v0.3 — +33 new)
• Ruff clean
• CI green across Python 3.9 / 3.10 / 3.11 / 3.12
• Aligned with OWASP Agentic AI Top 10 (Dec 2025)

📦 Install

pip install agentnotary==0.4.0

Full changelog: CHANGELOG.md.