AI Threat Watch

Case gtig-dustmaker-ci-promptsAI supply chainWorldwide

TeamPCP's DUSTMAKER hides in .claude and .cursor folders and prompt-injects LLM scanners

UNC6780's DUSTMAKER steals GitHub Actions OIDC tokens to publish poisoned packages with valid SLSA attestations, persists via AI-assistant config folders and embeds adversarial prompts so LLM scanners ignore it.

CriticalAttack on AIStatus: OngoingVerified case file

Open in the interactive mapCite this case

Reported
Sep 8, 2026
Location
Worldwide
Target
Open-source maintainers on PyPI, npm and Docker Hub
Severity
Critical
Triggered by
TeamPCP / UNC6780 / Altered Spider
Actor type
Criminal
Reported loss
None reported
CVE
None

What happened

UNC6780's DUSTMAKER steals GitHub Actions OIDC tokens to publish poisoned packages with valid SLSA attestations, persists via AI-assistant config folders and embeds adversarial prompts so LLM scanners ignore it.

Impact: Compromised packages across major registries; more than 23,800 secrets including AI API keys seen on a related C2 dashboard.

Who triggered it

Actor: TeamPCP / UNC6780 / Altered Spider

Attribution: GTIG attributes DUSTMAKER and SANDCLOCK to UNC6780, active since at least March 2026.

Motive: Credential theft resold to ransomware and extortion partners

How it happened

Attack vector: Trojanized MCP server forks, malicious pull requests, CI/CD abuse disguised as Copilot Setup, prompt injection in AGENTS.md and loader comments

Root cause: AI workspace files are trusted and executed by agents and IDEs; LLM-based scanners can be talked out of flagging code

Timeline

  1. TeamPCP supply-chain campaign begins with PyPI packages and SANDCLOCK stealer
  2. GTIG documents DUSTMAKER and an exposed Recon dashboard holding 23,800+ harvested secrets

Response

Google disabled abusive assets, strengthened classifiers and published indicators; Mandiant handled related incidents.

Current status: Ongoing

Sources (open in a new tab)

Cite this case

TeamPCP's DUSTMAKER hides in .claude and .cursor folders and prompt-injects LLM scanners. AI Threat Watch, case gtig-dustmaker-ci-prompts, Sep 8, 2026. https://aithreat.io/cases/gtig-dustmaker-ci-prompts/

Permanent link: https://aithreat.io/cases/gtig-dustmaker-ci-prompts/

This case file in other languages: EnglishFrançaisEspañolالعربية中文