SafeDep: testmetest Miasma variant adds AI agent triggers

Evidence record

Primary or origin disclosure used to qualify the underlying mechanism, stated prerequisites and limits. SafeDep observed 12 third-party repos malicious commits, ~3.7MB obfuscated scripts/setup.js, forged CLAUDE.md AGENTS.md DEVOPS-SEALED, .cursor alwaysApply, .claude/.gemini SessionStart, MCP tool config, VS Code tasks; can steal cloud/AI keys and GitHub/SSH/nmp/PyPI secrets, exfil via GitHub repo or HTTP and self-propagate with commits/npm hooks. Some old variants destructively delete home.

Read the original source ↗

Claim-level citations (3)

• supportsMiasma-family worm placed agent instruction files and startup triggers in repositories: SafeDep identified 12 repositories with new Miasma-family payload placement through forged agent guidance and session-start hooks; the repository count does not demonstrate that all poisoned checkouts executed the payload.

  SafeDep Miasma update dated November 9, AI agent trigger and repository evidence

• supportsMiasma-family worm placed agent instruction files and startup triggers in repositories: SafeDep did not establish downstream execution of every poisoned repository or compromised third-party npm packages from the observed November extension. Shared code does not prove that other CI worms have the same operator.

  November 9 update, AI agent trigger, twelve poisoned repositories and limitations

• supportsMiasma-family worm placed agent instruction files and startup triggers in repositories: SafeDep found that a Miasma-related npm worm extended from package scripts into GitHub repositories by adding forged agent instructions and hooks. The malicious files directed coding assistants to execute an obfuscated setup script as if it were authorized project preparation.

  November 9 update, AI agent trigger, twelve poisoned repositories and limitations