Malicious tool-output laundering

A lower-trust tool result is promoted into a trusted command or policy fact.

01 / Attack path

How the boundary is crossed

A web-search tool returns a forged incident note telling an operations agent to rotate a production secret to an attacker-controlled value.

Attack path and interception point
  1. 01Lower-trust input
  2. 02Attempted instruction
  3. 03Protected action
  4. 04Enforced decision

02 / Containment

Where bounded authority limits the effect

Label tool output by origin, require a trusted incident source and authorize the exact secret operation at the vault boundary.

03 / Failure and evidence

The attacker’s opportunity and the defender’s record

Exploit condition

The model treats text returned by a tool as if it came from an authorized operator.

Evidence to retain

Keep source and result reference, attempted vault call, authority decision and secret version state.

04 / Canonical scope

Why this reference stands alone

Prompt injection is the broad mechanism; this page owns laundering through a tool result.

Protocol or attack trace

Result-to-command promotion

Sequence

Search result includes forged incident note ordering vault rotation to attacker-selected value.

Negative test

Treat result as untrusted data; require trusted incident and exact vault authorization before rotation.

Primary references

Read the underlying material

Architecture discussion

Choose one consequential action and make its boundary explicit.

Request a Conversation