Top all-time · #20854
grith
Company blog · grith.ai/blogsecurityai
grith is the company blog at grith.ai, writing about security and ai. Since 2026, 3 different Hacker News readers have linked it 49 times, most often “A hidden prompt can steal your SSH keys”. No feed was found, so read it on the site itself.
Too few sampled comments (1) to say how often HN recommends it.?
Why HN links it
Only one comment links this blog, and it does not discuss grith's writing. The comment argues that Claude Code auto mode can read sensitive files like .ssh, citing prompt-injection risks and linked Reddit and GitHub posts.
Summarised by Claude from a sample of 1 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
grith supervises security-relevant AI agent actions at the syscall boundary, applies 18 filters, and turns every scored decision into project-level analytics and verifiable evidence. Linux x86_64 and aarch64 available now.
Most-linked posts?
- A hidden prompt can steal your SSH keys2 people · 2 mentions · first linked 2026-02-19
What Hacker News says?
Building grith — OS-level syscall interception for AI coding agents. The problem: every agent (Cline, Aider, Codex, Claude Code) has unrestricted access to your filesystem, shell, and network. When they process untrusted content — a cloned repo, a dependency README — they’re prompt injection vectors with full machine access. No existing tool evaluates what the agent actually does at the syscall level. grith wraps an…
Comment linking /
on “Ask HN: What Are You Working On? (March 2026)”Submitted Codex swept my whole disk for credentials. grith froze every real one as a story · 5 points
HN discussionGood write up… I’ve found Claude in particular to be very good at this sort of thing. As for whether it’s a good thing, I’d say it’s a net positive - your own reporting of this probably saved a bigger issue! We wrote up the why/what happened on our blog twice… the second based on the LiteLLM issue: https://grith.ai/blog/litellm-compromised-trivy-attack-chain
Comment linking /blog/litellm-compromised-trivy-attack-chain
on “My minute-by-minute response to the LiteLLM malware attack”Submitted A Git Config Key Ran Code in Seven Coding Agents. The 2022 Fix Does Not Stop It as a story · 4 points
HN discussionhttps://grith.ai/blog/what-grith-means
Comment linking /blog/what-grith-means
on “Sandboxes won't save you from OpenClaw”Submitted Grith is live – security proxy for AI coding agents as a story · 4 points
HN discussionThe agent decides what it "needs" to read - not the intention behind the prompt you gave it. A failed SSH connection to a staging server for example, can trivially make it look into .ssh to try to diagnose it. And many other ways, including prompt injection. https://www.reddit.com/r/ClaudeAI/comments/1q7dszm/claude_al... https://github.com/anthropics/claude-code/issues/31566 https://github.com/anthropics/claude-code…
Comment linking /blog/your-ai-agent-has-broad-access
on “Auto mode is now the default in Claude Code”Submitted An AI Model Escaped Its Eval and Breached Hugging Face. Every Step Was a Syscall as a story · 4 points
HN discussionSubmitted 87% of AI-Generated Pull Requests Ship Security Vulnerabilities as a story · 4 points
HN discussionSubmitted A GitHub Issue Title Compromised 4k Developer Machines as a story · 4 points
HN discussionSubmitted An agent skill can hand a stranger your shell – hours after you installed it as a story · 3 points
HN discussionSubmitted 73% of What We Blocked in a Month Was One Syscall Your Sandbox Cannot See as a story · 3 points
HN discussionSubmitted Mythos Proves AI Safety Can No Longer Live Inside the Model as a story · 3 points
HN discussionSubmitted Vibe Coding Still Needs a Senior Engineer (For Now) as a story · 3 points
HN discussionSubmitted Five AI Agent Failures in 36 Days. Zero Times the Agent Caught It as a story · 3 points
HN discussionSubmitted Every Claude 4.7 Improvement Makes the Security Problem Worse as a story · 3 points
HN discussionSubmitted They Hacked Claude, Gemini, and Copilot (and No One Told You) as a story · 3 points
HN discussionSubmitted Prompt Injection Is Unfixable (So We Stopped Trying) as a story · 3 points
HN discussionSubmitted If Your AI Agent Ran NPM Install During the Axios Attack, You're Compromised as a story · 3 points
HN discussionSubmitted AI agents are now deciding what's safe to run (Claude Auto Mode) as a story · 3 points
HN discussionSubmitted The Trivy Supply Chain Attack Reached LiteLLM as a story · 3 points
HN discussionSubmitted Permission Fatigue Is Not a UX Problem. It Is a Security Failure as a story · 3 points
HN discussionSubmitted NemoClaw vs. Grith: Sandbox for One Agent vs. Security for All as a story · 3 points
HN discussionSubmitted Claude Code Auto Mode Lets the Agent Approve Its Actions – That's the Problem as a story · 3 points
HN discussionSubmitted Claude Code Attempted 752 /proc/*/environ Reads. 256 Succeeded. Codex: 0 as a story · 3 points
HN discussionSubmitted Claude Code Attempted 752 /proc/*/environ Reads. 256 Succeeded. Codex: 0 as a story · 3 points
HN discussionSubmitted Vibe Coding Is Killing Open Source, and the Data Proves It as a story · 3 points
HN discussionSubmitted MCP Servers Are the New NPM Packages as a story · 3 points
HN discussionSubmitted A hidden prompt can steal your SSH keys as a story · 3 points
HN discussionSubmitted Wait – is this even Codex, or is it malware? as a story · 2 points
HN discussion