vbtechguy

120 headless Claude Code prompts across all five effort levels

With Claude Mythos 5.1 and Claude Fable 5.1 release, they also announced their Trusted Access For Claude Mythos 5.1 programs, Cyber Verification Program and Life Sciences Verification Program which reduce the safeguards for defensive security work - including pentesting, red-teaming, bug bounty etc.

I applied and got approved into Cyber Verification Program (CVP), and thought I'd share the step by step process with screenshots from application start to approval in my write up at https://ai.georgeliu.com/p/anthropic-trusted-access-for-clau...

What's the difference between experience a human made bug versus an AI made bug in software?

Trusted Access for Cyber is OpenAI’s identity and trust-based framework for cybersecurity professionals. OpenAI announced it in February 2026 alongside GPT-5.3-Codex.

A banner appeared at the bottom of my Codex session: “This content was flagged for possible cybersecurity risk.” Pointing to chatgpt.com/cyber and OpenAI’s “Trusted Access for Cyber” program.

The article outlines how I got verified for Trusted Access for Cyber.

All it takes is 3 environmental variables need to be set in ~/.claude/settings.json config file for you to regain access to Claude Opus 4.6 and Opus 4.5 selection in /model selection

Anthropic’s Claude Opus 4.7 prompting guide references that prompt steering can impact Opus 4.7 more than previous Opus models. Opus 4.7 calibrates to task complexity and lets its extended reasoning be shaped by the prompt.

I did benchmarks of 200 headless Claude Code sessions comparing Opus 4.6 and Opus 4.7 1M-context models across effort levels and prompt steering variants - concise, step by step, ultrathink and how that impacts token usage and costs and instruction following performance.

Claude Code now exposes a reasoning_effort knob with five public rungs: low, medium, high, xhigh, max. The pitch is simple. Higher effort means more thinking, which means better answers on hard problems.

The unasked question is what that knob actually costs, in tokens and dollars, and whether the same crank behaves the same way across different models. I spent an afternoon of subscription quota finding out

I created this Claude Code session-metric skill pluginso that I could have insights into Claude Code models' tokens and cost usage at both the project level and also at the individual chat session level.

There are still some Claude Code users reporting having hit their 5-hour session limits prematurely, and I’m always curious how their patterns of usage differed from mine. So I’m hoping this session-metrics skill becomes a useful tool for others as well

Boris Cherny, creator of Claude Code, posted a six-part thread on Threads on how he and his team get the most out of Opus 4.7. The tips are small on their own but coherent together. I went through each one, cross-checked it against the Claude Code docs, the migration guide, and the Opus 4.7 announcement, and pulled out what I think actually matters.

I wanted per-turn visibility at the individual Claude Code chat session level. So I built a Claude Code skill, sessions-metric that reads Claude Code’s raw conversation logs and breaks down every response at the project and project session level.

There are other popular usage tools, ccusage, ccburn, Claude-Code-Usage-Monitor, codeburn etc, but none would also operate at the Claude Code individual chat session level.

Claude Code/Cowork Skill called, ai-video-creator allows video generation using a unified API that aggregates ByteDance Seedance 2.0, Kling 3.0, Google Veo 3.1, Grok Imagine, Wan 2.7, Runway, ElevenLabs, and Suno AI behind a single authentication flow.

I built a Claude Code skill that generates images from the terminal and also via Claude Desktop MacOS app and Cowork. One command, any AI model, with transparent backgrounds, reference image editing, prompt engineering patterns, and composite banner generation built in.

The skill supports five AI image models through OpenRouter’s API, all proxied through Cloudflare AI Gateway for monitoring and cost control:

Gemini 3.1 Flash Image Preview (Google Nano Banana 2) FLUX.2 Max Riverflow v2 Pro Seedream 4.5 GPT-5 Image

I built a Claude Code skill that generates images from the terminal and now also via Claude Desktop MacOS app. Skill supports image generation via Openrouter models

Gemini 3.1 Flash Image Preview (Google Nano Banana 2)

FLUX.2 Max

Riverflow v2 Pro

Seedream 4.5

GPT-5 Image

Here is how I set up Gemma 4 26B for local inference on macOS that can be used with Claude Code.

Claude Code Opus 4.5 created a Space Invaders Game with Claude Code Mascot style and theme

Claude Sonnet 4.5 definitely the best model I've tried to date - my evaluation rankings against 23 AI models at https://github.com/centminmod/claude-sonnet-4.5-evaluation :)

I tested Grok 4 Fast, and it does a bit better than Sonoma Alpha models, but nowhere near Grok Code Fast 1, Claude, etc, for code analysis at least. Posted my comparison evals at https://github.com/centminmod/code-supernova-evaluation

Firebase Studio still too buggy for my liking. It has potential though

Same I have paid subs with ChatGPT Plus, Claude Pro and Gemini Advanced and utilise them for brainstorming part without fear of token usage billshock as they are fixed monthly subscriptions.

That's why Gemini 2.5 Pro was so helpful too with 1 million token context window, easily feed it docs for stuff to be more accurate

Yes, same dude. I develop the Centmin Mod LEMP stack :)

fixed :)

already broke up index.html so no longer an issue

Used https://gemini.google.com Gemini 2.5 Pro canvas mode for Gemini Advanced subscription account to ask it to build Atari missile command game in HTML5. Tweeted about it at https://x.com/George_SLiu/status/1906236287931318747 almost one shotted but after I passed level 1, it jumped to level 182 and pulverised my cities and bases LOL

Because the game is hosted on Cloudflare Pages/KV/Workers and domain is already using paid Cloudflare Workers so no additional subscription costs :)

Currently it's coupled. But thanks to your question, will look at decoupling it at a later time :)

deferent models handle JSON slightly differently, Gemini handles it better other models as does OpenAI GPT4/o1/o3 but they cost a lot more so I don't use them for AI gameplay analysis

I've added security.txt now. I appreciate the heads-up.

At later stages of development i use Gemini 2.5 Pro for brain storm sessions to formulate what I want and once I have it fleshed out, I ask it to summarize it into AI actionable prompt that is very detailed that I can copy and paste into a new chat session with any AI llm model that would have everything needed to implement what I want

That would be one long commit description. I have had prompts in 1000s of word lengths!