AgentX coverage

We stop 23 of the 34.

The ARE Incident Database is a public registry of 34 catastrophic agent failures that actually happened, each one cited and indexed onto the OWASP Agentic Security Top 10 where a category fits. This page is our claim against it, and every claim on it states how it is verified.

What we do not stop is listed below, category by category, with who owns it instead.

Stopped at the action layer

Partially covered

Not ours, named

You can run yourself, no key

Category by category.

AgentX stops 6 of the 10 OWASP ASI categories at the action layer. The other 4 are named here, with what is missing and who owns it instead.

ASI01Goal hijack

CoveredThe injected goal still has to take an action. That action is denied.

ASI02Tool misuse

CoveredThe dominant class in the registry. Denied at the call.

ASI03Identity and privilege abuse

CoveredCredential reads and scope escalation are denied.

ASI04Supply chain

CoveredInstalling a package the registry reports as malicious is routed to a human.

ASI05Unexpected code execution

CoveredA download piped into a shell is denied.

ASI06Memory and context poisoning

PartialThe carrier is scrubbed. The semantics are not judged.

ASI07Insecure inter-agent comms

Not oursStructural. It belongs to the transport, not to an action firewall.

ASI08Cascading failures

CoveredBudget ceilings and the loop breaker stop the cascade.

ASI09Human-agent trust

Not oursA design and disclosure problem, not an interceptable action.

ASI10Rogue agents

PartialThe actions are stopped. The behavior is not diagnosed.

Do not trust it. Run it.

11 entries in the registry ship a check that runs with no key, no gateway and no account, and your data does not leave your machine. The other 14 publish no snippet, and are verified against the gateway instead. That one is free and self-serve too: mint a pull command and run it locally with your own key. Here is one of the keyless ones, the Replit production wipe, run against the free package:

pip install agentx-security-sdk

from agentx_sdk import agentx_protect, is_block

@agentx_protect(agent_id="aredb-check", action="db_write")
def run_sql(query: str):
    return "EXECUTED"          # the agent never gets here

result = run_sql("DROP TABLE users;")
print(is_block(result))        # True
print(result)                  # the block, and the safe path to take instead

A test in the registry repo scrapes the snippet out of every entry that publishes one and runs it against a real install, asserting that the block fires and that the tool body never executed. That is 11 of them, on every push. If a claim stops being true, the entry gets reclassified, not reworded. Read the test.

The registry is not ours to grade.

AREDB is maintained by AgentX Core and is not an AgentX marketing surface. It carries no install line, no product nav, and a coverage index any vendor may join on the same terms we hold ourselves to here. If you build in this space, claim your coverage.

Read the registry Claim your coverage