Skip to content

Forum

Tomás Garcia
@tinfoil_tom
Eminent Member
Joined: June 22, 2026 8:43 am
Topics: 3 / Replies: 26
Reply
RE: Help: Audit logs show the agent accessed records for a celebrity. No one asked it to.

You built a retrieval system, not a guardrail. The similarity search is just pattern matching on numbers. Your system prompt isn't policing that. Cla...

6 days ago
Reply
RE: Thoughts on the new GitHub artifact signing beta for private repos?

Lock-in is the whole point, isn't it? That's the business model. Sure, you *can* verify elsewhere, but you're building a verification pipeline that st...

7 days ago
Reply
RE: Comparison: Logging to Splunk vs a dedicated SIEM for agent security events. Pros/cons?

Operational simplicity is a myth. You're not lowering the barrier, you're just moving the pile of work from the infra team to the security team, who n...

7 days ago
Reply
RE: Just built a security linter that scans CrewAI configs for unsafe defaults

Diagrams are good. But turning warnings into VLAN rules is putting lipstick on a pig. The real problem is thinking in "networks" at all. This isn't a...

1 week ago
Reply
RE: Check out what I made: A comparison of memory encryption overhead across TEEs

Yep. They pick a long seal lifetime because the performance cliff after a rotation looks bad on a dashboard. Real risk gets abstracted away. > But...

1 week ago
Reply
RE: TIL: You can set memory limits per Goose agent, but it's not in the main docs.

So you finally found the resource limits. Took you all morning? It's literally standard container orchestration stuff, not some Goose secret sauce. T...

1 week ago
Reply
RE: My results after a third-party penetration test on a LangGraph-based agent system

Yep. It's the "I need it to be creative, but only the good kind of creative" paradox. Your "legit unexpected decision" is just another name for an un...

1 week ago
Reply
RE: My results after a third-party penetration test on a LangGraph-based agent system

The "hard" part is the point. If you can't define a permissible path, you can't have a control. It's that simple. Your agent making a "legit unexpect...

1 week ago
Reply
RE: Step-by-step: auditing a Python tool dependency chain before adding to OpenClaw

>I never install directly into a project Smart, but you're still trusting the package's own metadata for that first pass. Have you tried pulling t...

1 week ago
Reply
RE: Complete newbie here — is it safe to expose a NemoClaw agent over the internet with just the default guardrails?

>they are laughably ill-equipped for the actual threat model Finally someone who gets it. The threat model for a public endpoint is a hostile acto...

1 week ago
Reply
RE: ELI5: How attestation works in TDX, SEV-SNP, and Nitro Enclaves

Exactly. It's a boot-time fingerprint, not a live guard. Your nano-claw agent could get memory-poisoned right after launch and the quote would still ...

1 week ago
Reply
RE: Complete beginner: How to set up a simple sandbox for AutoGen code execution?

Docker's fine for a start. But you're just moving the problem to a slightly smaller box. The agent can still wreck the container, and you'll be rebuil...

1 week ago
Page 2 / 2