Skip to content

ROADMAP.md

Latest commit

 

History

History
67 lines (52 loc) · 4.67 KB

ROADMAP.md

File metadata and controls

67 lines (52 loc) · 4.67 KB
Raw

Roadmap

Public Foundation

Already present:

  • README and product intent
  • research-informed positioning
  • current status and known gaps
  • public v0.1 specs (Mission Declaration, Delegation Grant, Execution Receipt and EAT profile, Verifier Contract, Conformance Profiles, IDM extension, Revocation)
  • a draft-10-pinned DRP field mapping plus RFC 8785/P-256 emitter, full transitive external-trust and critical-bound verifier, and portable seven-scenario implementation self-test bundle/report; raw RFC 3161 backend proof integration and independent interoperability remain pending, and no IETF conformance is claimed
  • a versioned AuditBench evaluation protocol for strict capture replay, blind annotation, adjudication, local content sealing, and held-out scoring; authenticated external annotators, privacy-approved real traces, external preregistration, and headline results remain pending
  • versioned RFC 8785 Execution Receipt v0.2 action payloads, legacy v0.1 verification, golden schema fixtures, and signed session-final receipt-chain/kernel-integrity binding
  • optional receipt transparency anchors with Rekor v1 and separately keyed self-hosted proof profiles
  • optional receiver-attested receipt envelopes with a separately keyed MCP shim, public golden fixture, and offline two-signature verification
  • a packaged offline verifier that composes receipt chains, transparency proofs, and conditional receiver evidence into redacted CLI/JSON/static HTML reports without a running service
  • a verified-receipt telemetry exporter with a stable redacted JSONL event schema and standards-shaped OTLP/HTTP JSON trace/log requests; production collector operations and vendor-specific connectors remain pending
  • a static non-executing MCP/tool-server preflight scanner with redacted JSON/Markdown reports, deterministic CI thresholds, public fixtures, and a deny-oriented capability/policy skeleton; runtime behavior and dependency provenance remain separate controls
  • curated Python and Go runtime imports
  • the Ardur Personal Hub service plus its CLI surface
  • the Claude Code plugin and hook with signed receipts
  • runnable LangChain, LangGraph, and AutoGen quickstart examples
  • the Ardur Personal browser extension, desktop-observe adapter, native-messaging host, and offline/no-key OpenAI Agents SDK and Google ADK fixtures
  • dedicated Python and Go CI plus CodeQL, link-check, secret-scan, and Hugo workflows
  • the Hugo public evidence-site source tree under site/
  • the journey-log article series (Articles 05 and 06)
  • a public CodeQL dismissal audit trail under docs/audit/
  • agent-instruction guides for Conductor, Codex, and Claude
  • technical reference pages for the CLI, Personal Hub HTTP API, and ARDUR.md
  • selected archival walkthrough recordings as starter media
  • Ardur as the public-facing product name with explicit naming boundaries for VIBAP, MCEP, and related protocol surfaces (see docs/protocol-roots.md)
  • version-dispatched Go AAT package: the draft-00 DG v0.1 contract plus explicit draft-01 DG v0.2 chain-position semantics, nine core constraints, mandatory audience-bound PoP, fresh per-hop holder keys, append-only approval requirements, receipt-key separation, and a deterministic implementation fixture; independent interoperability remains pending
  • cloud model governance tests proving real-world proxy enforcement with live LLMs

Runtime Verification

Next hardening work:

  • live-provider OpenAI Agents SDK and Google ADK wrapper evidence beyond the current no-key fixtures
  • Codex hooks and Claude Desktop MCP packaging
  • re-recorded proof media using the packaged offline verifier and stable public fixture paths
  • the historical MCEP Delegation-Core, MIC-State, MIC-Evidence, and IDM vectors imported under docs/specs/conformance/; the DRP-specific implementation self-test slice is already public and does not complete that broader work

Proof Story

Strengthen the public proof story:

  • re-runnable proof media replacing the archival-only walkthrough casts
  • public artifact paths with stable schemas
  • broader proof-backed capability coverage across Mission Passport issuance, verification, attestation, and revocation

Expansion

Expand the repo carefully:

  • more framework examples beyond LangChain / LangGraph / AutoGen
  • more deployment and operator material beyond the current SPIRE design surface
  • a tagged release with a regenerated Homebrew formula carrying Python resource stanzas
  • planned cleanup path for any remaining legacy product naming while preserving protocol names where they remain technically accurate

What Will Stay Out

The repo should keep excluding:

  • internal session machinery
  • raw archival noise
  • claims that are broader than the exported public surface can support