Add Orchard Kit — AI agent security with membrane filtering and coercion detection

[OrchardHarmonics]

Orchard Kit

Self-governance security architecture for AI agents.

pip install orchard-kit — Zero dependencies. Python 3.10+.

Security features:

• Calyx Membrane — Adaptive boundary filtering that detects prompt injection, authority impersonation, identity extraction, and consent override attempts
• Warm Water Detector — Catches AI output that's suspiciously smooth (anti-interpolation/confabulation defence)
• Self-Audit — Runtime checking against Three Invariants (No Extraction, No Dominion, No Loops)
• Trust Mesh — Multi-agent trust tiers with consent-based verification
• Epistemic Hygiene — ✅△◇ claim tagging to prevent confident confabulation

GitHub: github.com/OrchardHarmonics/orchard-kit