From ea9535ec052c67e9fb6d08eee5aaf329cb467cd6 Mon Sep 17 00:00:00 2001 From: Claude Bot Date: Fri, 24 Jul 2026 08:55:53 +0800 Subject: [PATCH] docs: clarify Trust Passport architecture in org README MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Trust Passport spec โ†’ wasmagent-protocol Trust Passport product โ†’ open-agent-audit agent-trust-infra = AgentBOM/MCP Posture validators + CLI --- profile/README.md | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/profile/README.md b/profile/README.md index 435ad0b..c7ca965 100644 --- a/profile/README.md +++ b/profile/README.md @@ -44,7 +44,7 @@ investment. | --- | --- | | [wasmagent-js](https://github.com/WasmAgent/wasmagent-js) | **Runtime** ยท Embedded agent runtime v1.x โ€” WASM kernels (QuickJS, Pyodide, Wasmtime, Remote), MCP gateway + attestation, AEP emitter, capability manifests; adapters for A2A, AG-UI, AI SDK, and Claude Agent SDK. This is where signed evidence is born. | | [symkernel](https://github.com/WasmAgent/symkernel) | **Verification** ๐Ÿšง ยท Go symbolic verification backend โ€” cel-go lightweight rules, wazero Wasm sandbox hard-isolation, Z3 SMT proofs; HTTP service consumed by wasmagent-js and wasmagent-py. "Proving it ran correctly" is the mission itself. | -| [agent-trust-infra](https://github.com/WasmAgent/agent-trust-infra) | **Trust & compliance** ยท AgentBOM, MCP Posture, and Trust Passport spec, reference impl, and CLI; EU AI Act Annex IV compliance mapping (draft, deadline **2026-08-02**). | +| [agent-trust-infra](https://github.com/WasmAgent/agent-trust-infra) | **Trust tooling** ยท AgentBOM and MCP Posture validators, `@wasmagent/trust-cli` developer CLI, EU AI Act Annex IV compliance mapping (draft, deadline **2026-08-02**). Trust Passport spec โ†’ [wasmagent-protocol](https://github.com/WasmAgent/wasmagent-protocol). | > **`wasmagent-py`** *(planned)* joins the spine as the Python runtime sibling โ€” > same AEP schema, Criterion/ConstraintIR protocol, and symkernel adapter โ€” so @@ -122,8 +122,10 @@ GPU training jobs: it reads PyTorch Flight Recorder dumps, builds a cross-rank PROV-DM causal graph, and produces tamper-evident `EpochEvidenceBundle` records, enabling tensor-origin tracing and deterministic replay. -`agent-trust-infra` layers on trust artifacts โ€” AgentBOM, MCP Posture, and Trust -Passport โ€” giving every agent run a machine-readable identity and policy posture. +`agent-trust-infra` produces trust artifacts โ€” AgentBOM and MCP Posture โ€” giving every +agent run a machine-readable identity and policy posture. Trust Passport spec lives in +`wasmagent-protocol`; the Trust Passport product (issuance, signing, revocation) lives in +`open-agent-audit`. An EU AI Act Annex IV compliance mapping (Article 11 / Annex IV, effective 2026-08-02) is in draft, covering 20 of 29 Annex IV sub-items. @@ -154,7 +156,7 @@ sustained track record. - **Verification** โ€” `symkernel`, cel-go rules, wazero sandbox, Z3 SMT integration (Go) - **Pipelines** โ€” `trace-pipeline` / `evomerge` (measurement trust, admission, training audit) - **Training evidence** โ€” `wasmagent-train-replay`, PyTorch Flight Recorder, PROV-DM, GPU training causal graphs -- **Trust artifacts** โ€” `agent-trust-infra` (AgentBOM, MCP Posture, Trust Passport, EU AI Act mapping) +- **Trust tooling** โ€” `agent-trust-infra` (AgentBOM/MCP Posture validators + CLI); Trust Passport spec โ†’ `wasmagent-protocol`; Trust Passport product โ†’ `open-agent-audit` - **Audit product** โ€” `open-agent-audit` / Trustavo (evidence reports, Cloudflare Workers) - **Evaluation** โ€” `fresharena` (dynamic, verifiable, adversarial evaluation; paper preparation) - **Adapters** โ€” OpenTelemetry GenAI, Langfuse, LangSmith ingestion