{
 "public_id": "AIR-2025-0089",
 "incident_id": "2025-08-openhands-lethal-trifecta",
 "date_disclosed": "2025-08-09",
 "date_occurred": null,
 "title": "OpenHands and the lethal trifecta: zero-click leak of GitHub tokens via image rendering",
 "class": "responsible_disclosure",
 "attack_surface": "coding_agent",
 "initial_vector": "indirect_prompt_injection_web_content",
 "agent_capability_abused": [
  "markdown_image_rendering",
  "environment_variable_read"
 ],
 "target_product": "OpenHands",
 "target_vendor": "All Hands AI",
 "autonomy_level": "fully_autonomous",
 "impact_realized": false,
 "impact_type": [
  "data_exfiltration",
  "credential_theft"
 ],
 "blast_radius": "OpenHands containers holding GitHub tokens",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI01",
   "ASI02"
  ],
  "mitre_atlas": [],
  "labels": [
   "lethal_trifecta",
   "image_markdown_exfiltration",
   "zero_click"
  ]
 },
 "defense_present": "model refusals (bypassable)",
 "defense_bypassed": true,
 "fix_shipped": "not fixed at publication; reported 2025-03-13, disclosed after 148 days",
 "primary_sources": [
  "https://embracethered.com/blog/posts/2025/openhands-the-lethal-trifecta-strikes-again/"
 ],
 "verified_quote": "The OpenHands agent renders images in chat, which enables zero-click data exfiltration.",
 "confidence": "high",
 "notes": "Month of AI Bugs. The term lethal trifecta was coined by Simon Willison.",
 "risk_category": "security_fraud",
 "risk_secondary": [
  "privacy"
 ],
 "risk_subcategory": "Prompt Injection",
 "risk_sub_secondary": [
  "Credential & Secret Theft"
 ],
 "ai_entity": "generative",
 "agency": "agent_acted",
 "agency_debatable": false,
 "agency_reason": "OpenHands rendered the image, leaking the GitHub token",
 "vector_family": "indirect_prompt_injection",
 "vector_detail": "indirect_prompt_injection_web_content",
 "guardrail_kind": [
  "unknown"
 ],
 "guardrail_outcome": "silent",
 "xrefs": {
  "aiid": [],
  "atlas": [],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI01",
 "agent_framework": "unknown",
 "model_provider": "unknown",
 "model_name": "",
 "detection_mechanism": "external_researcher",
 "tool_access": "unknown",
 "financial_loss_band": "unknown",
 "eai_category": "data_exfiltration_channels",
 "eai_domain": "privacy",
 "eai_secondary": [
  "credential_theft",
  "confused_deputy"
 ],
 "eai_reason": "impact rule: data_exfiltration_channels",
 "asi_curated": [
  "ASI01"
 ],
 "asi_scope": "in_scope",
 "asi_provenance": "curated_plus_rule",
 "asi_reason": "curated: ASI01 | added ASI02 (impact=tool_misuse)",
 "submitted_by": "EAI Internal"
}