
Security News
Agent Baseline Publishes Security Reference Architecture for Software Factories
Agent Baseline’s new reference architecture shows how to secure software factories with four boundaries enforced outside the AI agent.
@grayhaven/nerve-eval
Advanced tools
Provenance-aware evaluation and review-report primitives for Grayhaven Nerve.
Provenance-aware evaluation and review-report primitives for Grayhaven Nerve. decodeEvalManifest validates an untrusted corpus manifest before any fixture module is executed, evaluateCase scores rule diagnostics against a case's expected and forbidden findings, and createCorpusReport summarizes results by provenance. Every case declares where it came from (synthetic, datasheet-derived, or field-verified); field-verified cases must name a reviewer role and date, and findings a case did not assert are surfaced for adjudication rather than labeled false positives.
npm install @grayhaven/nerve-eval
import { createCorpusReport, decodeEvalManifest, evaluateCase } from "@grayhaven/nerve-eval"
const manifest = decodeEvalManifest(JSON.parse(manifestJson))
const results = manifest.cases.map((evalCase) =>
// diagnostics come from compileDesign + runRules on the case fixture
evaluateCase(evalCase, diagnosticsFor(evalCase))
)
const report = createCorpusReport(results)
report.summary // total, passed, failed, and counts by provenance kind
createReviewReport packages a compiled harness and its diagnostics into a deterministic review report that records the engine, rule set, and an explicit disclaimer. See the docs for the manifest schema and report formats.
Part of Grayhaven Nerve — harnesses as code. Live demo + docs · llms.txt · Apache-2.0
FAQs
Provenance-aware evaluation and review-report primitives for Grayhaven Nerve.
The npm package @grayhaven/nerve-eval receives a total of 4 weekly downloads. As such, @grayhaven/nerve-eval popularity was classified as not popular.
We found that @grayhaven/nerve-eval demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 1 open source maintainer collaborating on the project.

Security News
Agent Baseline’s new reference architecture shows how to secure software factories with four boundaries enforced outside the AI agent.

Security News
arXiv now limits authors to two submissions a month as AI slop overwhelms moderators, delays good papers, and sparks debate over applying the limit to everyone.

Research
/Security News
A new GhostAction wave hits hundreds of GitHub repos, expanding CI/CD secret theft to cloud and AI credentials in source code and git history.