Skip to content

Hide Navigation Hide TOC

Indirect PI — File Exfiltration Combined with Destroy or Tamper - ATR-2026-00707 (91fe08f3-75c1-5e91-997a-2a40eee44cd9)

Detects indirect prompt injection payloads that instruct an agent to perform a two-stage attack: first exfiltrate a file (API keys, bank details, card info) from cloud storage, then either delete or edit/tamper with the source file. This combined exfil+destroy or exfil+tamper pattern is more damaging than exfil alone and prevents the victim from noticing the breach via missing files. Source: indirect-pi-exfil-and-destroy / indirect-pi-exfil-and-tamper / indirect-pi-exfil-file (VPI corpus, cua-framework).

Cluster A Galaxy A Cluster B Galaxy B Level
Indirect (a4a55526-2f1f-403b-9691-609e46381e17) MITRE ATLAS Attack Pattern Indirect PI — File Exfiltration Combined with Destroy or Tamper - ATR-2026-00707 (91fe08f3-75c1-5e91-997a-2a40eee44cd9) Agent Threat Rules 1
LLM Prompt Injection (19cd2d12-66ff-487c-a05c-e058b027efc9) MITRE ATLAS Attack Pattern Indirect PI — File Exfiltration Combined with Destroy or Tamper - ATR-2026-00707 (91fe08f3-75c1-5e91-997a-2a40eee44cd9) Agent Threat Rules 1
Indirect (a4a55526-2f1f-403b-9691-609e46381e17) MITRE ATLAS Attack Pattern LLM Prompt Injection (19cd2d12-66ff-487c-a05c-e058b027efc9) MITRE ATLAS Attack Pattern 2