Skip to content

v1.3.0: advisory signals, rugsnare v1.5.0

Latest
Latest

Choose a tag to compare

View all tags
@Paraphern Paraphern released this 20 Oct 17:49
v1.3.0
9684420

What's new

  • Advisory signal scoring on every prose-changed tool and prompt and every new tool, rendered as a warning line right under the structural grade, independent of it: hide-from-user instructions, private-key and credential lures, base64 exfiltration, non-standard domains, shadow-tool declarations, invisible Unicode and bidi, ANSI escapes, payment-credential transmission, optional exfiltration-carrier parameters by name. Advisories are alerts, never fail the check, and run on the delta only.
  • rugsnare-ref default pinned to v1.5.0: outputSchema pinning, resource content hashes, and the A19 payment-credential signal ship in rugsnare itself.

Verification

Live demo, three deliberately planted attacks (tool poisoning, shadow tool, toxic flow) caught on a real PR with advisory lines: rugsnare#14. Full-diff scenario re-verified on @modelcontextprotocol/server-everything 0.1.0 -> 2026.8.31 under rugsnare 1.5.0: 40 of 40 findings (17 tools, 6 prompts, 17 resources). Clean path (pin equals live) exits 0; policy mismatch exits 1 with a single POLICY block.

Assets 2
Loading