01 / 05PLANNED
Threat Intelligence
Understand the attack, affected agent workflows, evidence, indicators, and mitigations.
Prompt injection · Tool poisoning · Data exfiltrationStructured threat records with references and update dates
Get ZTAgent ↗
Menu
03 / KNOWLEDGE BASE PAID PRODUCT · IN DEVELOPMENT
From threat to tested response.
News tells you what happened. ZTAgent Security Intelligence is being built to show what the threat means for your agents, which controls apply, and how to test them.
Paid access and downloads are closed while reviewed content and billing are prepared.
FREE GitHub / secure your agentsFREE News / stay informedPAID Knowledge / stay protected
FIVE CONNECTED CATEGORIES
Intelligence you can act on.
These are the planned content areas. Listings below appear only when an editor has reviewed and published them.
02 / 05PLANNED
Detection Rule Packs
Deploy reviewed ZTAgent detections with applicability notes, versions, tests, and known limits.
Prompt injection · MCP security · Secrets protectionVersioned, downloadable rule packs when published
03 / 05PLANNED
Security Policy Packs
Translate an organization’s risk boundaries into decisions for agent and tool actions.
MCP Zero Trust · Browser agent · Coding agentConfiguration examples and deployment guidance
04 / 05PLANNED
Attack + Test Library
Reproduce an attack and check the expected detection and security response.
Tool abuse · Indirect injection · Data leakageTest cases linked to the threat and applicable controls
05 / 05PLANNED
Standards + Guidance
Map primary standards and advisories to controls a team can evaluate in its own environment.
OWASP · NIST · MITRE ATLAS · CISAStandard → threat → rule → policy → test
THE PAID VALUE
One incident. Four useful answers.
A proposed workflow for turning a primary source into something a team can deploy and test. This is a product example, not a published pack.
- 01 / THREATHow does a poisoned tool response redirect the agent?
- 02 / DETECTIONWhich suspicious instructions or tool metadata can be identified?
- 03 / POLICYWhich untrusted actions should require a deny or approval?
- 04 / TESTDoes the protected workflow produce the expected result?
Rules and policies need testing against your agent’s tools, permissions, and legitimate workflows. No single control stops every attack.
EDITOR REVIEWED / LIVE CATALOG
Published intelligence.
Only reviewed entries and packs from the publication database appear here. Public summaries and original source links stay visible.
No reviewed paid releases yet.
We are preparing the first linked threat, control, and test material. Browse the free security news or request early access.
Start with open source. Follow the news. Apply reviewed controls when the Knowledge Base opens.