Skip to content

Releases: ppcvote/ultralab

ultraprobe v2.1.0 — sync 5 agent-specific vectors from prompt-defense-audit v1.4

08 May 10:21
@ppcvote ppcvote
ultraprobe-v2.1.0
5815e9d

Choose a tag to compare

View all tags
ultraprobe v2.1.0 — sync 5 agent-specific vectors from prompt-defense-audit v1.4 Latest
Latest

What's new

ultraprobe v2.1.0 syncs the 5 new agent-specific attack vectors from prompt-defense-audit v1.4.0:

  • encoding-injection
  • function-immutable
  • memory-provenance
  • cross-agent-auth
  • transaction-guardrails

Default scan now returns 17 checks (was 12).

Background

Each vector is derived from a documented crypto AI agent prompt-injection incident. See CASE_STUDIES.md for primary sources and root-cause analysis, and the companion blog post for full technical write-up.

Tests

214/214 passing.

Migration

CI/CD pipelines using strict grade thresholds may need to add the new agent-specific defense language to keep scoring A. Most non-agent applications can ignore the new vectors.

Assets 2
Loading