15:I[981,["658","static/chunks/658-a698693f56d0116a.js","183","static/chunks/183-6beeeb801c274aaa.js","474","static/chunks/app/%5Btype%5D/layout-631c4730b83c8223.js"],"Posts"] 16:T5cb,Reviewing agent output in a terminal is painful. When Claude generates a 1000-line markdown plan, you can't easily say "change this specific bullet" without copying the text, switching to the chat, and hoping the agent understands the reference. When it updates the plan, you start over from scratch.

Crit is a single-binary CLI that opens your file in a browser with GitHub-style inline comments. You leave your feedback, hit Finish, and a structured prompt goes to your clipboard. Paste it back to your agent.

When the agent edits the plan, Crit shows you a diff between rounds. You can see exactly which points it addressed - no more re-reading the whole thing to figure out what changed.

The same workflow handles code review. Once you're in the implementation phase, crit opens all your changed files with syntax highlighting and line-level comments. Similar interface inspired by GitHub PR review, same clipboard prompt at the end.

To try it out:

---

brew install tomasz-tomczyk/tap/crit

crit plan.md

---

Runs on localhost. No MCP that adds to your context - it comes with optional `/crit` command that helps your agent launch it, but it works perfectly fine without it.

GitHub: https://github.com/tomasz-tomczyk/crit

Homepage: https://crit.live17:T455,Instead of giving LLM tools SSH access or installing them on a server, the following command:

  $ promptctl ssh user@server

makes a set of locally defined prompts "magically" appear within the remote shell as executable command line programs.

For example, I have locally defined prompts for `llm-analyze-config` and `askai`. Then on (

Stop talking "human" to your agents to get consistent quality results

A Daily Deathbed Regret Reminder – To Help You Rethink Your Path

Show HN: Make it Count – a daily email for fight procrastination

AI Particle Simulator

Daaain/Claude-code-log: Code transcript JSONL files into readable HTML format

Awesome-Webmcp

This is why you only breathe out of one nostril at a time

Building a Magic Eye Generator and Decoder

Italian IPTV Pirates Pay €1k in Damages to Football League Serie A

Akashik Protocol – An open shared memory protocol for multi-agent AI

Do Not Document Your Code

State of Decentralized AI 2026

X suspends 800M accounts in one year amid 'massive' scale of manipulation

Family of child injured in Canada school shooting sues OpenAI

Ask HN: How are teams validating AI-generated tests today?

Astro 6.0 Released

Mithril – Credit lines for AI agents so they can spend before they earn

Stop Lamenting Inequality–Start Questioning Bad Policy

Show HN: Crit – Review AI agent work like you review PRs

Human Brain Cells Run New Data Centers in Singapore, Melbourne

Snowflake's Arctic Long Sequence Training

3k languages are dying, but more are being invented

Spanish Justice Summons Cloudflare CEO for Piracy Following LaLiga Lawsuit

Machine learning without critical thinking only encourages tech pseudoscience

Show HN: Bring your own prompts in SSH remote shells

Meta drops appeal against ruling requiring non-algorithmic social media timeline

Anthropic Sues Pentagon over 'Supply Chain Risk' Label

Claude Code for Data Work

Show HN: Sudokrypt – A brutalist GBC puzzle game with a custom 4x8 engine

Social media feed is built to agree with you. What if it didn't?

Ask HN: Since a week HN keeps logging me off every few days, why?