Skip to content

Changelog

Release history for SecureVibe. Every version is an Ed25519-signed release (see Signing); full notes on GitHub Releases.

v0.9.0 — 2026-06-21 · Initial public release (ShieldNet360)

First public, open-source release of SecureVibe under ShieldNet360. A deliberate pre-1.0 cut: the product is feature-complete across the PREVENT → DETECT → ENFORCE → LEARN lifecycle and ships with reproducible, CI-gated benchmarks; the remaining supply-chain trust automation (CodeQL, OpenSSF Scorecard, SLSA provenance, SBOM, reproducible-build verification) is tracked on the roadmap for v1.0.

Capabilities at launch

  • PREVENT — 29 signed SKILL.md knowledge docs injected into AI coding assistants (Claude Code, Cursor, Copilot, Codex, Windsurf, Cline, Devin) via MCP, so secure code is written at generation time.
  • DETECT — 4 deterministic scanners (secrets, dependencies, Dockerfile, GitHub Actions) over a 2,022-entry, web-cited malicious-package canon across 10 ecosystems, plus OSV and 27 Sigma rules.
  • ENFORCE — gate blocks insecure diffs in CI (severity floor, SARIF → GitHub Code Scanning); reusable GitHub Action and pre-commit hook.
  • LEARN — skills-check contribute: a signed local overlay flywheel (you → team → org) with peer-to-peer submit / verify / import.
  • Trust — Ed25519-signed releases and data manifests; manifest-checksum, dist-drift, and scanner-eval-drift CI gates; offline-first, keyless.
  • Docs — Material site with an in-browser WebAssembly Playground, per-role guides, a Threat-Intelligence browser, compliance-coverage matrices, and reproducible benchmark methodology.
  • Eval — a 110-fixture, judged prevention-lift harness; the first trustworthy measured prevention-lift (+7.4 points with the skill in context, +10.0 points with the scanner tool, false-positives flat — see Benchmarks).

Installs: curl | sh, Homebrew tap, go install, and npx.