feat: add multi-vendor agent collaboration demo

Demo showing trust-gated handoffs between OpenAI, Anthropic, and
Google agents with cross-vendor audit trail and trust scoring.

Closes #161

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Author: mosiddi

examples/07-multi-vendor-collaboration/README.md

@@ -0,0 +1,116 @@
+# Example 07: Multi-Vendor Agent Collaboration
+
+This example demonstrates **trust-gated handoffs** between agents from different AI vendors, orchestrated through AgentMesh's governance layer.
+
+## Scenario
+
+A customer support pipeline where three vendor-specific agents collaborate:
+
+```
+Customer Query
+      │
+      ▼
+┌─────────────┐   trust handoff   ┌─────────────┐   trust handoff   ┌─────────────┐
+│  Agent A     │ ───────────────▶  │  Agent B     │ ───────────────▶  │  Agent C     │
+│  (OpenAI)    │                   │  (Anthropic) │                   │  (Google)    │
+│  Customer    │                   │  Research    │                   │  Summary     │
+│  Queries     │                   │  Analysis    │                   │  Generation  │
+└─────────────┘                   └─────────────┘                   └─────────────┘
+      │                                  │                                  │
+      └──────────────────────────────────┼──────────────────────────────────┘
+                                         ▼
+                              ┌─────────────────────┐
+                              │   AgentMesh          │
+                              │   Trust & Governance │
+                              │   ─────────────────  │
+                              │   • Trust scoring    │
+                              │   • Audit trail      │
+                              │   • Policy engine    │
+                              │   • Capability scope │
+                              └─────────────────────┘
+```
+
+## What It Shows
+
+| Feature | Description |
+|---|---|
+| **Cross-vendor trust** | Agents from OpenAI, Anthropic, and Google verify each other before handoffs |
+| **Adaptive trust scores** | Trust increases on success, degrades on failure or policy violations |
+| **Trust-gated delegation** | Tasks only route to agents meeting minimum trust thresholds |
+| **Cross-vendor audit trail** | Every handoff logged with vendor, timestamp, trust score, and outcome |
+| **Untrusted agent handling** | Demonstrates graceful degradation when a vendor's trust drops below threshold |
+
+## Running the Demo
+
+No API keys required — all vendor agents are mocked.
+
+```bash
+cd examples/07-multi-vendor-collaboration
+python demo.py
+```
+
+### Expected Output
+
+The demo runs three rounds:
+
+1. **Round 1** — All agents trusted, full pipeline succeeds
+2. **Round 2** — Agent C (Google) produces poor results, trust degrades
+3. **Round 3** — Agent C's trust drops below threshold, handoff is denied and fallback activates
+
+After all rounds, a full cross-vendor audit trail is printed.
+
+## Key Concepts
+
+### Trust-Gated Handoffs
+
+Before delegating work to another vendor's agent, AgentMesh verifies:
+
+```python
+handshake = TrustHandshake(min_score=700)
+result = handshake.verify_peer(target_agent)
+if result.trusted:
+    delegate(task, target_agent)
+else:
+    activate_fallback(task)
+```
+
+### Adaptive Trust Scoring
+
+Trust scores update after every interaction:
+
+- **Success** → trust increases (e.g., +50 points)
+- **Slow response** → minor trust penalty (e.g., −30 points)
+- **Policy violation** → major trust penalty (e.g., −200 points)
+- **Failure** → significant trust penalty (e.g., −300 points)
+
+### Cross-Vendor Audit Trail
+
+Every handoff is recorded with full provenance:
+
+```
+[2024-01-15 10:00:01] HANDOFF agent-a-openai → agent-b-anthropic | trust: 850 | status: APPROVED
+[2024-01-15 10:00:02] HANDOFF agent-b-anthropic → agent-c-google  | trust: 800 | status: APPROVED
+[2024-01-15 10:00:03] TASK_COMPLETE agent-c-google | trust_delta: -150 | reason: poor_quality
+```
+
+## Extending This Example
+
+- **Add more vendors** — Register additional agents with different vendor tags
+- **Custom trust policies** — Adjust `MIN_TRUST_SCORE` or add vendor-specific thresholds
+- **Real API integration** — Replace mock agents with live OpenAI/Anthropic/Google calls
+- **Persistent audit** — Export the audit trail to a database or SIEM
+
+## Production Considerations
+
+| Concern | Recommendation |
+|---|---|
+| Trust persistence | Store trust scores in a durable backend across restarts |
+| Vendor SLAs | Set per-vendor trust thresholds matching contractual SLAs |
+| Audit compliance | Export audit trail to immutable storage for regulatory review |
+| Fallback chains | Define fallback agents for every vendor in the pipeline |
+
+## Learn More
+
+- [Trust Architecture](../../docs/TRUST.md)
+- [Governance Model](../../GOVERNANCE.md)
+- [AgentMesh Security](../../SECURITY.md)