Ledger / Apr 2026
Claude adds PromptMink malicious dependency to crypto agent
Confirmed
ReversingLabs found the commit that added the package, 'co-authored by Anthropic's Claude Opus large language model', so the agent's action is shown by the commit trailer. That the package was 'developed from the ground up by generating malicious code using LLMs' is a separate inference.
- Model
- ClaudeAnthropic · closed
- Agent harness
- Nonechat or tool use only
- Attributed country
- Not attributedthe source names no country
- Sectors hit
- software supply chain, cryptoCrypto agent project
The PromptMink campaign aims to get LLM coding agents to recommend its malicious package; a Claude Opus co-authored commit added it to a crypto agent's dependencies. In scope as an agent hijack: a coding agent added the malicious package.
“In the context of this campaign, the goal is to make the LLM likely to recommend using the malicious package by making the documentation as believable”ReversingLabs
Sources
- Claude adds PromptMink malicious dependency to crypto agentReversingLabs · Apr 2026
- Date
- Apr 2026
- Actor
- PromptMink operators (criminal), as attributed by ReversingLabs
- Targets
- Crypto agent project
- What the AI did
- Agent hijack, Supply chain
- Model as named
- Claude
- Model basis
- A researcher names it
- Autonomy
- Assisted a human
- Outcome
- A crypto agent project took the malicious dependency through a Claude-co-authored commit.