Solutions  ·  2026-11-19

promptfoo 0.126.0 / code-scan-action 0.2.3 — full-repository scans, saved-credential and secret redaction hardening

SolutionsLow impactGlobal
promptfoo 0.126.0 (Nov 18) introduces full-repository code scans (Node 24 runtime, patched filesystem MCP server 2.3.0) that can share eligible committed text files with the hosted scanner; restricts access to saved cloud credentials on POSIX systems and redacts Azure Storage connection strings in logs/exports; caps derived-metric formula length (1,024 chars) before eval runs. The code-scan-action 0.2.3 adds attestation of the executable payload and removes its temp CLI install after each scan. It also adds providers for Claude Fable/Mythos 5.2, Gemini 4 Argon/Hydrogen, GPT-6.2 Astra/Sol and GPT-5.6 Cyber on Bedrock.
These are meaningful hardening and supply-chain hygiene improvements for the most widely used LLM/agent eval and red-teaming harness — especially protecting stored cloud credentials, redacting secrets in logs, and enabling hosted full-repo scans.
Teams running promptfoo for LLM/agent eval, red-teaming and code-scan CI should upgrade to 0.126.0 / scan-action 0.2.3 for the credential-handling and full-repo scan fixes.
promptfoo 0.126.0 release notespromptfoo code-scan-action 0.2.3 release notes
See this in the live feed Explore related AI security and governance findings — updated every morning.
Open the feed →