DOP 376: Secrets Management for AI Agents

November 11, 2026 Episode 376

Show Notes

#376: How many secrets can your coding agent reach right now? You gave it a token so it could get work done. It might have passed that token to every agent it spun up. Somewhere along the way, you stopped reading the permission prompts and started clicking allow.

At the same time, hardcoding secrets in our code is still the most common answer for how secrets are managed. Developers get paid to close tickets, not to lower risk, and the agents just copy whatever they find in the repo.

In this episode, we speak with Dwayne McDaniel, Principal Developer Advocate at GitGuardian, about managing secrets when agents are the ones using them, and why the better answer might be no secrets at all.

Frequently Asked Questions

Why don't developers use secrets managers?

Developers skip secrets managers because they are paid to ship fast, not to reduce risk, according to Dwayne McDaniel of GitGuardian on DevOps Paradox episode 376. Security tools ask people to go slowly and correctly, and when lowering risk competes with production flow, production flow wins every time. Viktor Farcic agrees from the developer side: he supports more security as long as it is not more painful, and wants tools that work almost transparently.

Why do AI coding assistants hard code credentials?

AI coding assistants hard code credentials because they follow whatever template or existing pattern they find, and many templates still put the credential directly in the code. Dwayne McDaniel of GitGuardian points out on DevOps Paradox episode 376 that this works, just not safely. Viktor Farcic adds that agents copy what a repository already does, so once a project starts hard coding secrets, the agent will not push back and fix the earlier work.

Is a database connection string a secret?

A database connection string is a secret if it contains a credential, says Dwayne McDaniel of GitGuardian on DevOps Paradox episode 376. Darin Pope raised the case, arguing that a leaked connection string points attackers at a real endpoint and is just as bad as the credential itself. Dwayne treats credentials, secrets, and certificates as the same thing: if a string unlocks access to data, it needs protecting, and a vault can hold any variable you consider risky.

Does storing a password in a secrets manager make it safe?

Storing a password in a secrets manager does not make it safe. Darin Pope notes on DevOps Paradox episode 376 that a password stored six years ago and never touched has probably leaked somewhere else by now. Dwayne McDaniel of GitGuardian, who strongly supports secret managers, recommends governance policies that flag secrets unused for a year and automate the follow-up. The bigger cultural problem, he says, is creating access without thinking about how it gets revoked.

How should AI agents be identified?

Dwayne McDaniel of GitGuardian argues on DevOps Paradox episode 376 that an AI agent is a workload and should be identified the way workloads already are, using SPIFFE and its runtime, SPIRE. Attestation at creation proves which process, chip, and firmware spawned it, and a fresh SVID can be issued every five minutes. He calls SPIRE an eight-year-old, free and open source technology, and disagrees with those in the identity community who say agents need a new approach.

What should you do when a credential leaks?

Rotation is still the answer when a credential leaks, says Dwayne McDaniel of GitGuardian on DevOps Paradox episode 376. If the secret is already called from a vault, rotate it there, and a blue-green deployment should keep anything from going down. The catch is that this is a human speed response. He cites a CrowdStrike case where only 27 seconds passed between an attacker breaking through and exfiltration starting, because the attack harvested every credential it could reach.

What is the DevOps Paradox podcast?

DevOps Paradox is a weekly podcast co-hosted by Darin Pope and Viktor Farcic, covering DevOps, platform engineering, and modern software delivery. Episode 376 features Dwayne McDaniel of GitGuardian on secrets management for AI agents: why credentials still end up hard coded, how workload identity frameworks such as SPIFFE apply to agents, and why leaked secrets now get exploited at machine speed. Every episode page carries audio, video, and a full transcript.

Episode Transcript

Read the Full Transcript

Guests

Dwayne McDaniel

Dwayne McDaniel is a Principal Developer Advocate who has been on a mission to “help people figure stuff out” for over a decade. At GitGuardian, he specializes in secrets security and non-human identity governance across cloud and DevOps environments. A frequent speaker at events like DevOpsDays and BSides, he helps security and engineering teams better understand complex issues.

Hosts

Darin Pope

Viktor Farcic

Viktor Farcic is a member of the Google Developer Experts and Docker Captains groups, and published author.

His big passions are DevOps, Containers, Kubernetes, Microservices, Continuous Integration, Delivery and Deployment (CI/CD) and Test-Driven Development (TDD).

He often speaks at community gatherings and conferences.

He has published DevOps Paradox and Test-Driven Java Development.

His random thoughts and tutorials can be found in his blog The DevOps Toolkit.