What happened
promptfoo released version 0.121.16 on June 16, 2026, adding two features: (1) provider support for Anthropic's Claude Fable and Mythos models in the red-teaming framework, enabling security teams to run adversarial evaluations against these frontier models; and (2) publication of all four promptfoo red-team skills to the Claude Code marketplace, making automated red-team capabilities directly accessible from Claude Code workflows. A companion 0.121.17 release (same day) fixed Docker/Python pinning issues and matcher bugs.
Why it matters
Fable and Mythos are the most security-capable Claude models (Mythos was specifically cited in AWS Continuum and CISA/G7 AI SBOM guidance as a capability inflection point). Adding them to promptfoo's red-team provider list means security teams can now formally evaluate and jailbreak-test these models before enterprise deployment. The Claude Code marketplace integration brings red-teaming into developer CI workflows.
Applicability
Security teams evaluating Anthropic Fable/Mythos for enterprise deployment should use promptfoo 0.121.16+ to run red-team sweeps before go-live. AI red-teamers using Claude Code can now invoke promptfoo skills directly from their coding environment.