Banks: User-controlled prompt input can be parsed as privileged chat messages

[ Why it matters ]Priority signals

• Public proof-of-concept exploit exists (CISA SSVC)

[ Description ]Official description

Banks generates meaningful LLM prompts using a simple template language. Prior to 2.5.0, Banks Prompt.chat_messages() attempts to parse every line of rendered template output as ChatMessage JSON. When an application renders untrusted data and passes the returned ChatMessage objects to an LLM provider, attacker-controlled JSON can cross the prompt boundary and become a system, assistant, or tool message because ChatMessage.role accepts arbitrary strings. This can override application instructions, alter the intended prompt structure, or confuse downstream tool and message handling. This issue is fixed in version 2.5.0.

[ Fixes ]Fixes and patches

• banks 2.5.0Fixed version · GitHub Advisories · pip package banks: upgrade to 2.5.0 or later (affected <= 2.4.5)

[ Packages ]Affected open-source packages

• banks pipAffected <= 2.4.5 · fixed in 2.5.0 · GHSA-hmq2-7hp6-7crh

[ References ]References

• github.com/masci/banks/commit/02172b816fb84f6a824cc09a8aca7416f53c12cb
• github.com/masci/banks/pull/78
• github.com/masci/banks/releases/tag/v2.5.0
• github.com/masci/banks/security/advisories/GHSA-hmq2-7hp6-7crh