←
Insecure Agents Podcast377 posts
Insecure Agents Podcast avatar

Insecure Agents Podcast

@insecureagents

AI engineers and security practitioners listen to us to learn how to give their agents the security they need to reach full autonomy and capability.

◷ Joined June 2025▧ 191 Photos and videos
Insecure Agents Podcast avatar
Insecure Agents Podcast@insecureagents·21h

Roll Your Own Agent Harness: Building Daiku and Personal Agents with Tejas Kumar "Let me send a prompt here, 'Delete everything.' That's what I did on stage. And it actually just did it." @TejasKumar_, AI Engineer and author of Fluent React, talked to us at @aiDotEngineer CODE in SF about his AIE CODE workshop, building agent harnesses from scratch, and why "can the agent call this tool" and "should it call this tool right now" are two different questions. We get into: > Why banks, insurers, and hospitals are building their own harnesses instead of using off-the-shelf coding agents > The guardrails that came after rm -rf: allow lists, ask-first tools, and Face ID before sensitive actions > Getting the right human in the loop, not the whole team > Context caching, the piece most self-built harnesses are missing > Daiku, a personal agent harness that isn't locked to one model vendor TIMESTAMPS (00:40) The workshop: from an empty while loop to a coding agent harness (01:45) Rolling your own harness for air-gapped enterprise infrastructure (02:40) Why regulated industries can't just use off-the-shelf harnesses (03:40) Can you call the tool vs should you call it right now (04:10) "Delete everything" on stage, and the guardrails that came next (04:50) The right human in the loop (05:50) Context caching and why self-built harnesses overpay for tokens (07:20) The Resy incident: an agent that tried too hard (07:55) AAuth and budgets for downstream resources (09:50) Keeping credentials out of the agent and the deployment (11:15) The power plug you can pull on a rogue agent (12:05) Daiku: one computer every one of your agents can drive (13:05) The agent that signed Tejas up for Shark Tank (15:45) Switching models mid task when you run out of usage

17:35
Insecure Agents Podcast avatar
Insecure Agents Podcast@insecureagents·Nov 5

What Changes When Agents Write the Code: Gabriel (Yalo) on Software Factory Security "I haven't seen in the past months a pull request from a human at Yalo for something related with UI." Gabriel Brolo, Staff Security Engineer at Yalo and a @snyksec Ambassador, joined us live at @AISecSummit to explain how you secure a software factory when agents write the code. His answer: defense in depth, re-secured constantly, because one layer isn't enough. We get into: > Scoping authorization on APIs and services that were built before agents > Agent firewalls at the edge, backed by asynchronous LLM judges > Scanning every skill for credentials and exfiltration paths > Why blocking a misbehaving agent at runtime is a hard problem TIMESTAMPS (00:00) Gabriel, Yalo, and teaching compilers (01:00) Two years as a Snyk design partner (02:40) The Secure Software Factory reference architecture (03:00) An agent that builds agents: securing it is securing the factory (03:40) Role-based access and authorization on pre-agentic services (04:10) Agent firewalls: redacting PII and blocking topics in microseconds (05:00) Asynchronous LLM judges when the edge isn't sure (05:40) Red teaming against prompt injection and the OWASP LLM Top 10 (06:30) From drag-and-drop flows to Yalo Code (09:00) Responding to a misbehaving agent, and why live blocking is still open (12:50) No human UI pull request in months (15:40) Scanning skills for credentials and exfiltration (16:40) Re-securing the factory as models change and zero days land weekly (21:20) Automated pentests after every deploy: one scan doesn't work

23:55
nader dabit avatar
⟳ Insecure Agents Podcast retweeted
nader dabit@dabit3·Nov 3

S̶o̶f̶t̶w̶a̶r̶e̶ ̶e̶n̶g̶i̶n̶e̶e̶r̶i̶n̶g̶ Security engineering

Insecure Agents Podcast avatar
Insecure Agents Podcast@insecureagents·Oct 29

Julian Bright demoed a travel agent and used AAuth to verify the mission as the agent worked.

6:55
Insecure Agents Podcast avatar
Insecure Agents Podcast@insecureagents·Oct 22

We had great conversations last night about agentic runtime security and intent based auth at AARM intercept night. It was our pleasure to help host and look forward to the next event!

Post attachmentPost attachment
Insecure Agents Podcast avatar
Insecure Agents Podcast@insecureagents·Oct 21

Governance Is the Accelerator: Josh Woodruff (CSA) on Getting Agents Out of Pilot Purgatory "Zero Trust taught us to check who you are and what you can reach on every request. But agents need one more check on top of that. What are you doing right now?" Josh Woodruff, Co-Chair of the Zero Trust Working Group at @cloudsa, Founder and CEO of Massive Scale AI, and creator of the Agentic Trust Framework, joined us to explain why treating every AI agent like a new hire that earns its autonomy is what lets the business move faster with agents. We get into: > The five questions you have to answer about every agent, and the intern to principal autonomy ladder > Why agents running under a human's identity turn every kill switch into a work stoppage > Why intent is a useful signal > Why many smaller scoped agents is better than one monolithic one TIMESTAMPS (00:38) Helping leaders say yes to agents and keep the power to stop them (01:29) The book, the Agentic Trust Framework, and donating it to CSA (04:01) ATF and Agent Baseline land in the same place (05:12) Why every agent needs its own identity and behavioral baseline (09:07) Inventory, access reviews, and the 20-year-old shared service account (14:35) Human sponsors, tool scopes, and the roster as a forcing function (19:36) Intent-based auth, AAuth missions, and the deleted inbox (21:44) The Resy ban and the $1.4 million floor cleaner order (24:59) ATF and AARM: earned autonomy meets runtime action control (30:48) Fine-grained kill switches and microagents (32:45) Tokenomics and why constraints are rocket fuel (39:52) Reversibility and the never list (43:11) Secure action, not just access, and the Stop Rogue AI bill

46:07
Ian Livingstone avatar
⟳ Insecure Agents Podcast retweeted
Ian Livingstone@ianlivingstone·Oct 13

Today we're publishing the first draft of the Agent Baseline Software Factory Reference Architecture! An opinionated blueprint for securing increasingly autonomous software factories, built with folks across the community. It's open for feedback and contributions! agentbaseline.org/reference-…

Post attachment
Insecure Agents Podcast avatar
Insecure Agents Podcast@insecureagents·Oct 13

Join us tonight in New York right after @aiDotEngineer to discuss it at Secure Software Factory Show and Tell with @ianlivingstone, @AhmadNassri, @ealeyner, @DickHardt, @braintrust, @SocketSecurity, @Docker, @KeycardAI Join us at Elsie Rooftop 🗽🥂: luma.com/keycard-rg1c