From a senior to future senior Hackers

Why Lockfile Pinning Alone Won't Catch Hallucinated Imports from Coding Agents
Featured

Why Lockfile Pinning Alone Won't Catch Hallucinated Imports from Coding Agents

Lockfile pinning alone won't catch hallucinated imports from coding agents because it validates declared dependency versions, not whether a package exists, is maintained, or was invented; teams need registry checks, package-age signals, and CI gates.

pr0h0•
ai-coding-agentsdependency-securitylockfilesci-cdsupply-chain

Latest Posts

View all posts →

Latest Videos

View all videos →
Detecting a Credential Stealer in npm and pnpm Dependency Trees

Detecting a Credential Stealer in npm and pnpm Dependency Trees

A poisoned release can run code in your build before your first import ever loads. This walkthrough audits npm and pnpm dependency trees for install-time credential stealers: enumerate lifecycle hooks with npm query, trace which parent dragged a package in, treat pnpm's ignored build scripts warning as a triage queue, and shrink the blast radius by denying dependency script execution and removing long-lived cloud credentials from the build environment. The case study is the reported poisoned Microsoft Agent Framework release on PyPI, which matters for JavaScript because the failure mode is the same: install-time code execution inside a build environment that holds ambient cloud identity. Lockfile pinning stops drift, but it does not stop a maintainer publishing a tarball you then choose to install. The demo is defensive and lab-scoped, using a synthetic fixture that writes a timestamp file and nothing else. No working exploit, no real credentials. Full write-up: https://hackyjs.com/posts/detecting-a-credential-stealer-in-npm-and-pnpm-dependency-trees

supply-chain-securitynpmpnpm
Discord Moved Desktop and Mobile to Rust and GPUI: What That Actually Tests

Discord Moved Desktop and Mobile to Rust and GPUI: What That Actually Tests

Discord's reported move of its desktop and mobile clients to Rust and GPUI is a framework story more than a Discord story. This video walks through what the report actually says — one aggregated news item, a Google News RSS redirect, no engineering detail — and what it leaves open: which platforms shipped, whether GPUI now targets iOS and Android, whether the Electron build is gone, and whether any performance numbers exist at all. The technical core is the layer swap. Electron isn't a UI toolkit; it's roughly fifteen subsystems shipped as one download. Auto-update and code signing, crash reporting and minidump symbolication, an accessibility tree for arbitrary widgets, web content embedding, mobile platform shells — none of that comes with GPUI out of the box. Chromium also handed the web team sticky positioning, scroll anchoring, overflow clipping, subpixel rounding and a hundred layout edge cases their CSS quietly depended on. GPUI has been pressure-tested against exactly one product: the Zed editor, a text grid with a command palette and a file tree. A chat client stresses different surfaces — virtualized message lists with variable-height content, a composer with mentions and emoji, presence-heavy sidebars, voice and video. Whether GPUI's abstractions survive that is the event worth following. The claim being defended here is that the rendering swap is the easy thirty percent. The remaining seventy percent is platform work that doesn't demo well and decides whether the project ships on schedule. Confirmed, inferred and speculative are kept apart throughout. Every numeric block in the original write-up is a labelled output template, not a measurement, because the client is closed source and the rewrite is unconfirmed. Full post: https://hackyjs.com/posts/discord-moved-desktop-and-mobile-to-rust-and-gpui-what-that-actually-tests

rustgpuidiscord
Scaling Agents on Kubernetes with acpx and ACP: Isolation, Routing, and Cost per Agent

Scaling Agents on Kubernetes with acpx and ACP: Isolation, Routing, and Cost per Agent

A 55-second walkthrough of how acpx and ACP package agent runtimes on Kubernetes: why the unit of work is a session rather than a request, where the gateway belongs, what has to live outside the pod, and why a prompt is not safe to retry blindly. What the video covers: - The ACP wire exchange: session/new, then session/prompt scoped by sessionId - Why a stdio JSON-RPC protocol needs a gateway inside each pod - Conversation, workspace and credential state that must survive pod eviction - The retry rule: only retry before the first session/update for a turn - Why scaling granularity is the session, not the request Honest caveat from the post: it has no access to the acpx runtime or a recording of the talk, so every load and cost figure is a placeholder with a stated method, not a benchmark. Run the probes yourself before you size anything. Full article: https://hackyjs.com/posts/scaling-agents-on-kubernetes-with-acpx-and-acp-isolation-routing-and-cost-per-agent

kubernetesai-agentsagent-infrastructure