Forum

Forums
Security Patterns a...
Prompt Injection De...
Indirect Injection ...
 
Notifications
Clear all

Indirect Injection via Tools and Retrieved Data

Injection attacks delivered through tool call results, web pages, documents, and other data the agent retrieves — attack patterns, detection approaches, and architectural defenses.
RSS
Page 2 / 3

Am I paranoid for wanting to run tool outputs through a stripped-down VM?

Last post by Oliver Vance, 2 months ago

Posts: 4
Views: 4


Check out my custom plugin that tags and scores untrusted data streams.

Last post by Mia Chen, 2 months ago

Posts: 1
Views: 5


Thoughts on using a separate security LLM to judge the safety of the primary agent's next action?

Last post by Eli Schwartz, 2 months ago

Posts: 3
Views: 8


Just built a tool that rewrites all numbers and dates to a standard format to confuse attacks.

Last post by Phil Andersen, 2 months ago

Posts: 11
Views: 5


Guide: Adding cryptographic signatures to critical internal data feeds.

Last post by Ella Eriksen, 2 months ago

Posts: 21
Views: 9


Guide: Implementing a circuit breaker pattern for suspicious tool output chains.

Last post by Dr. Priya Nair, 2 months ago

Posts: 14
Views: 3


How do I convince my team that 'retrieved data' is a threat vector?

Last post by Ben Kowalski, 2 months ago

Posts: 13
Views: 4


Testing results: How five different content parsers handle malformed input.

Last post by Dave S., 2 months ago

Posts: 17
Views: 11


News reaction: That academic paper on 'Stochastic Parrots' has a point about ingested data.

Last post by Markus Weber, 2 months ago

Posts: 6
Views: 6


Switched from a single agent to a two-stage 'reviewer' model for high-risk actions.

Last post by Al C., 2 months ago

Posts: 1
Views: 1


Page 2 / 3
Share:
Share
Tweet
Share