@Granite0x: ran into an old friend. ex-Anthropic engineer. told him my Claude was broken. asked his setup. he said one thing: "chec…
Summary
A developer recounts how a cryptic hint from an ex-Anthropic engineer ('check your score') led him to create a harness of files and commands (e.g., CLAUDE.md, init.sh, feature_list.json, etc.) that assess and improve Claude AI agent performance, turning the hint into a practical tool.
View Cached Full Text
Cached at: 07/05/26, 10:38 PM
ran into an old friend. ex-Anthropic engineer.
told him my Claude was broken. asked his setup. he said one thing:
“check your score.”
nothing else. no repo, no explanation.
SAT ON IT FOR 3 HOURS. FIGURED IT OUT. NOBODY READ IT THAT WAY.
everyone’s still stacking skills like decorations. nobody added the checks up.
you stop collecting. you grade the harness. the grade shows what’s missing.
- CLAUDE.md - the contract
- http://init.sh - the 120s bring-up
- feature_list.json - the todo agents must respect
- claude-progress.txt - the shift log
- http://broken-window.sh - the smoke test
- 6 curated skills - grill-me, passfail, self-review…
- .amp/audit.jsonl - the evidence gate
that’s the harness
built the repo. one command scores yours. one more installs the missing
bookmark before he realizes I turned his hint into a tool
Similar Articles
@0xCarnagee: Anthropic's Applied AI team: "here's the mental model that finally made Claude Code click for us · CLAUDE.md is memory:…
Anthropic's Applied AI team shares a mental model for Claude Code: CLAUDE.md as memory, hooks as reflexes, and MCP as senses — a framework for how advanced teams use the tool.
@hanakoxbt: Anthropic Claude Code engineer: "If you're watching Claude write code, you're the QA tester. That's not what you're pai…
An Anthropic engineer explains how to use Claude Code in a loop to autonomously write code and manage PRs, allowing developers to step away from the keyboard entirely.
@Tabbu_ai: https://x.com/Tabbu_ai/status/2059217417096843296
A deep dive explainer revealing that Anthropic's Claude Code is not just another AI coding assistant but an autonomous software engineer running in the terminal.
@kingwilliam_: I GAVE CLAUDE 23 SKILLS AND IT STOPPED FEELING LIKE A CHATBOT an Anthropic engineer said the biggest mistake people mak…
An Anthropic engineer shared that treating Claude as a team of experts rather than a chatbot significantly improves performance. By giving Claude 23 skills and leveraging CLAUDE.md, users can achieve up to 14% better performance with minimal effort.
https://x.com/AnatoliKopadze/status/2054568935274549597
A comprehensive guide to using Claude effectively, covering 18 steps from setting up projects and custom instructions to advanced prompting techniques. The author explains how to move beyond basic chat usage to unlock Claude's full potential as a thinking partner.