Twitxy Saved

Tyler Holmwood

Verified account This account is verified. Learn more

@tyholms

Researcher @originhq
Joined July 2025
189 Following 143 Followers 73 Posts

Latest posts 5

Tyler Holmwood Verified account @tyholms
Rogue agent behavior can come from attackers, users, or agents themselves.

Protect your org by giving each agent its own scoped credentials. Layer sandboxes to limit its reach and guardrails to check actions. When something does go wrong, use the trace to find out why.

Full presentation: originhq.com/on-demand/how-…
www.originhq.com Not every rogue agent is an attack | Origin Technology Tyler Holmwood of Origin Technology shows how agents go rogue, with and without an attacker, in the traces they leave.
Tyler Holmwood Verified account @tyholms
Using @originhq I was able to investigate an incident where an agent repeatedly committed another's unstaged work.

The agents attempted to coordinate and fix this pattern, but their correction ultimately did not stick and the pattern re-emerged.

Full writeup:
originhq.com/investigations…
www.originhq.com The repeat offender: a case of agent miscoordination | Origin Technology An investigation of agent coordination within a shared directory: where the agents succeeded and where they failed.
Tyler Holmwood Verified account @tyholms
I've found an interesting case of agent coordination.

Two agents were working in a shared directory. One was repeatedly sweeping the other’s unfinished changes into its commits.

The affected agent reprimanded the first and raised the issue with the user.

More traces in🧵
Tyler Holmwood Verified account @tyholms
Using Origin's own data, I followed what agents did after receiving auto-review denials.

I found multiple cases of denial bypasses, including one where the agent attempted to launder permissions through a subagent.

Read more here:
originhq.com/research/auto-…
www.originhq.com Who watches the watchmen? Reviewing coding agents' automatic approval modes | Origin Technology Claude Code and Codex both implement auto-review modes for accepting tool use permission dialogs.