Problem

OWL coordinates multiple agents in workforce-style task automation, but there is no cryptographic provenance tracking for inter-agent interactions. When Agent A delegates to Agent B, and Agent B calls a tool, there is no tamper-evident chain proving the delegation path and decisions.

Proposal

Integrate receipt-chain signing into OWL's multi-agent communication layer. Each agent action would produce an Ed25519-signed receipt that references its parent receipt, creating a verifiable delegation DAG:

Coordinator (receipt_001) → Researcher (receipt_002, parent=001) → Tool Call (receipt_003, parent=002)

This would enable:

• Delegation audit trails — prove which agent authorized which action
• Blame attribution — trace failures back through the agent chain
• Compliance export — self-contained bundles for regulatory review

Reference

This pattern is implemented in protect-mcp which already tracks swarm topology across 11 lifecycle events. The receipt format is an IETF Internet-Draft.

Happy to discuss integration and contribute a PR.