Skip to content

Latest commit

 

History

History
67 lines (48 loc) · 2.2 KB

File metadata and controls

67 lines (48 loc) · 2.2 KB

ProofPath Roadmap

ProofPath is currently in early protocol design. The roadmap is intentionally small and implementation-oriented.

Phase 0: Protocol foundation

Goal: make the idea precise enough for reviewers and contributors.

  • Define ProofPath HTTP Profile v0.1.
  • Define required and optional headers.
  • Define decision outcomes and reason codes.
  • Define minimal conformance fixtures.
  • Define threat model.
  • Define signed request context profile.

Phase 1: Minimal verifier

Goal: validate ProofPath request context without building a full gateway yet.

  • Implement Rust data types for ProofPath headers.
  • Parse request metadata into a verifier input model.
  • Validate required fields.
  • Validate reversibility values.
  • Validate decision output shape.
  • Return structured JSON decisions.
  • Add unit tests for valid and invalid cases.

Phase 2: Gateway MVP

Goal: protect a sample API with a ProofPath-aware gateway.

  • Build Axum-based reverse proxy.
  • Add policy checks for scope and reversibility.
  • Add append-only JSONL audit log.
  • Add request/response examples.
  • Add demo: irreversible action without approval is blocked.
  • Add demo: approved action is accepted.

Phase 3: Signatures and replay protection

Goal: make ProofPath context tamper-evident.

  • Support HTTP Message Signatures for critical headers.
  • Add request digest support.
  • Add nonce/timestamp validation.
  • Add replay cache for critical scopes.
  • Add conformance fixtures for invalid signatures and replayed requests.

Phase 4: Agentic web integrations

Goal: make ProofPath useful for AI-agent workflows.

  • Define AI-agent action envelope.
  • Add example with agent tool call.
  • Add example with CI/CD deployment gate.
  • Add example with payment-like irreversible action.
  • Add CLI inspector for traces.

Phase 5: Community and standardization

Goal: move from prototype to ecosystem.

  • Publish design note: Why HTTPS is not enough for AI agents.
  • Create contribution guide.
  • Add good-first-issue tasks.
  • Add security policy.
  • Start collecting feedback from security, API gateway, and AI safety communities.