Supply Chain Security, in plain English.

JSON Parsing Library Vulnerabilities You Should Know About

JSON is the lingua franca of APIs, but the libraries that parse it have had serious security issues. Here is what to watch for in your stack.

Webpack vs Rollup vs esbuild: A Security Comparison

Choosing a bundler is usually about speed and features. Here is how Webpack, Rollup, and esbuild compare on the dimension that matters most for supply chain security.

SEC Form 8-K Item 1.05: Two Years of Enforcement Lessons

Two years into mandatory cybersecurity incident disclosure, the SEC has issued comment letter sweeps and settled enforcement actions. Here is what filers got wrong.

Inside PyPI Project Quarantine: How the Reversible Takedown Workflow Has Performed Since Launch

PyPI's Project Quarantine status, introduced in August 2024 and used roughly 140 times in its first year, replaces irreversible deletions with a reversible hidden state. Here is how the workflow operates and how to consume the signal.

IBM Cloud Code Engine: A Supply Chain Defender's Walkthrough

Code Engine abstracts away Kubernetes for Knative-style serverless workloads on IBM Cloud. The supply chain story is different from what most defenders bring from AWS or GCP.

Refusal Rate Analysis: Griffin AI vs Mythos

A security AI that refuses too often is useless. One that refuses too rarely is dangerous. Griffin AI publishes calibrated refusal benchmarks; Mythos does not.

Path Traversal: Griffin AI vs Mythos

Path traversal is the vulnerability class that punishes lazy analysis. Framework-specific path normalisation, OS-dependent separators, symbolic link resolution, and archive extraction all hide exploitable gaps behind code that looks defensive. Griffin's engine resolves path operations with actual semantics; Mythos reads the variable name and calls it a day.

Griffin AI vs Inflection Pi for Security Assistance

ToxicSkills: When Claude Skills Become a Malware Distribution Channel

Snyk's ToxicSkills research found prompt injection in 36% of Claude skills tested and 1,467 malicious payloads. The SKILL.md trust model is the structural issue.