Skip to content

Security: jeremylongshore/intent-eval-dashboard

Security

SECURITY.md

Security Policy

Reporting a Vulnerability

Report security issues privately via GitHub Private Security Advisory or email jeremy@intentsolutions.io with subject prefix [SECURITY intent-eval-dashboard].

Please include:

  • A description of the issue
  • Steps to reproduce
  • Affected versions / commits
  • Any proof-of-concept code (if applicable)

We aim to acknowledge within 72 hours.

Supported Versions

intent-eval-dashboard follows semantic versioning. Pre-1.0.0 means surface is unstable and only the latest release is supported for security fixes.

Scope

In scope for this repository:

  • Source code and CI workflows in this repo
  • The deployed surface at labs.intentsolutions.io
  • The published dashboard-render/v1 predicate URI (when sequenced; lives at evals.intentsolutions.io/dashboard-render/v1, NEVER at labs.*)
  • The retraction protocol + retraction/v1 predicate

Out of scope (report to the relevant repo):

  • @intentsolutions/core schema/validator bugs → intent-eval-core
  • Audit-harness gate bugs → intent-audit-harness
  • Skill-eval issues → j-rig-skill-binary-eval
  • Rollout-gate Action bugs → intent-rollout-gate

Cryptographic posture

  • Evidence Bundles ingested into this dashboard are DSSE-wrapped, in-toto Statement v1, Rekor-anchored
  • Every ingest verifies signature + Rekor inclusion proof + schema before render (DR-035 § 4 B1 binding)
  • The dashboard itself sequences "sign-your-own-homework" attestation to v0.2.0+ pending second independent verifier (DR-035 § 4 B3)
  • Storage is content-addressed; upstream tag-deletion or force-push does not invalidate dashboard's local copy

What is NOT a vulnerability

  • Stale data displayed with a visible stale_since badge (this is the correct behavior — see DR-035 § 4 C4)
  • A retracted bundle showing as a tombstone page instead of 404 (correct behavior per DR-035 § 4 B4)
  • The absence of an aggregate "PASS%" headline (intentional per DR-035 § 4 C3)
  • The absence of basicauth on the public origin for operator views (correct per DR-035 § 4 C1; operator views are tailnet-only)

There aren't any published security advisories