📰 Dev.to · jaryn

Last Tuesday a worker refused to talk to Postgres. I did what every on-call engineer does without...

The patch looked harmless. A reviewer asked an agent to summarize a PR. The model got git diff...

Last week a colleague pasted a production config.yml into an AI chat, asked why the connection kept...

Three weeks after a team wired a free model allowance into their demo, the compliance review landed....

MiniMax H3 is the hot model in my group chat this week. Should we switch the coding agent to it? I...

A silent regression hit us after what looked like a routine cost optimization: someone pointed our...

An AI coding agent in your dev environment holds three things at once: your source code, your...

Last month I watched a teammate paste a stack trace into a hosted AI assistant. The trace contained...

A coding assistant proposes pip install requests-utils-pro at 09:41. By 09:43 the dependency is in a...

"A YAML risk policy and authorization fixtures keep GitHub Issues agent controls from becoming a permission bypass."

Turn a health-data connection into explicit assets, trust boundaries, abuse cases, and negative consent tests before launch.

Turn the July 21 Hugging Face incident into enforceable containment boundaries and regression fixtures.

Build a URL-fetch policy that validates every resolved address and every redirect hop instead of trusting the first URL.

A concrete least-privilege gate for evaluating a trending local-first code graph without turning repository context into tool authority.

A security readiness checklist for evaluating K3 open weights before letting them into your inference pipeline.

Treat metadata removal as one upload-security check, not proof that an image contains only image data.

Threat-model repository review instructions after GitHub expanded Copilot code-review customization.

Preserve raw request bytes, reject replay, and prove that semantically identical JSON is not signature-identical.

Turn Kimi K3's announced scale, context window, and planned weights release into an evidence-based self-hosting decision.

Turn CodeQL's new AI prompt-injection detection into a stable repository security contract with positive, negative, and SARIF assertions.

A small concurrent test exposes whether refresh-token rotation has atomic replacement, bounded reuse handling, and safe token-family revocation.

Convert a new defense-in-depth roadmap for increasingly capable agents into repository-level tests teams can run before granting tools and credentials.

A BYOK form is a trust-boundary decision. Audit endpoint identity, credential scope, runtime delegation, and egress before an agent can send prompts.

A practical outbound-data threat model and test procedure for coding agents with filesystem and network access.