WILL NESS

I built a variant of @mattpocockuk's grilling skill dedicated to frontend and it has improved how I build new apps and components. The general idea: 1. Use /grilling and /prototype as a base 2. Tell Claude to build 5 WILDLY different prototypes 3. Tell Claude to include a picker that lets you switch between each variant live 4. Each round you select your favorite(s) + leave feedback, and Claude will walk down each branch of the design tree, helping you zoom in on your desired design And THEN, I went and added it to /wayfinder, so whenever I make a new map and there's novel frontend work, a ticket is created specifically referencing that /grilling-frontend-prototyping needs to be invoked. This will not be the last time I build a cool skill and add it to Wayfinder; this is a very powerful pattern for planning work. You can find my skill here: github.com/will-ness-ai/s…

@plainionist Conspiracy theory: 1m context window is a pricing strategy more than a utility.

Serious question: Do we really need bigger context windows? I am moving in the opposite direction: restart the agent team after every ticket to reduce token usage. 🤷‍♂️

I posted about this earlier today: When responding to agents, ramble your response as you read. Just dictate your full thoughts as you respond to an agents questions as you read it, not after. It's ok if it's ugly. It's ok if you change your mind mid ramble.

@jackfriks @alperortac come plug your idea!

if i wanted to let my fiancee vibe code an app in pure english, whats the best setup so she can make a full app with user login and a database without ever doing anything but typing english?

Thank you for the response That kind of works, kind of doesn't. My two concerns are: (this is all AFAIK, you probably know better than me lol) 1. this PreToolUse hook with additional context is going to occur AFTER the LLM has already returned the Edit(file, content), so the issue is... The model has already emitted the Edit/Write without the 'custom rule' loaded. additionalContext will result in another turn from the LLM to fix whatever wasn't followed in the rule which has cost, speed, and context bloat implications. I could use PreToolUse with permissionDecisionReason but it has the same downsides. Also, I would need to implement a sentinel strategy tracking which of these 'custom rules' have already been loaded in the session to ensure future Edits/Writes don't re-load this 'custom rule' Although, now after thinking through this, even if rules had Tools in the frontmatter like I am requesting, they'd still have this same issue... We don't know what the model will do until after it emits the tool call in which its already too late. I guess my 'ideal' solution would be: A) Model emits Edit(file, content) B) Client sees that a Edit rule is applied, replaces Edit(file, content) from context window with the Rule content C) Model will likely emit Edit(file, content) again but without the original in the context window (yay, context savings!). From the context windows perspective it just happened to load the rule right before the Edit. D) The client wouldn't load the rule again for the rest of the session ^ This is something I cannot do with hooks in the current harness (AFAIK I can't surgically edit the context window?) 2. The main issue I am working around here is I am the AI DevX Engineer for my team and there's already a lot of friction getting people to contribute to skills, rules, etc.. I am looking for a less convoluted way for people to contribute :) I can use a skill for an adapter for now but its still friction.

@WillNessAI @bcherny @trq212 Thanks for the feedback! This doesn't answer your q but if you want something deterministic now, you can also use hooks. E.g. use a PreToolUse hook, set the matcher to the tool calls (Edit|Write), a conditional for the file path, and print your standards as additionalContext

@bcherny @trq212 I love the idea of Claude Rules, but they are unusable for my team. Rules are too broadly applied, I care about two dimensions: - Path (you already have) - Tools (you don't have) (what the agent is going to do) I want API_RULES.md (containing my coding standards) frontmatter to be: --- paths: - api/**/*.ts tools: - Edit - Write --- This means I wouldn't worry about the 'skill activation' problem, without muddying my context window for exploration tasks.

Pro-tip for my fellow dictators: dictate your response as you read, not after. (voice typist, not political leader) I used to read an entire assistant response from my agent (claude code) and then start dictating but its hard to address every point the agent has made in its response. Especially during grilling sessions agent responses can get really long and contain many possible things to correct or respond to. Dictating as I read ends up looking something like: "Yes I like point 3. No, I don't want to do strategy A because it'd break the current user experience. Yes, but we should do it this way instead of the way you suggested. Hmm, I like option 4. Okay, agreed." whereas without this I find that I read it all and just go: "okay lgtm" because I am lazy and don't want to go back and read individual points. This does lead to the case sometimes where you end up changing your mind for previous points, in which I use vim mode to surgically remove sentences, or just tell claude I changed my mind.

@troels_a @mikker @petersuhm I'll check out Yap to see if I could integrate it as the context layer.. right now I'm mixing Mongo for structured data and S3 for files and context

Nice! I haven't seen this before, very cool. I specifically *do* want a frontend. I use coding agents to maintain three surfaces for my database/API: - web app - mobile app - CLI (for agents) So I can access and modify my data via my computer, my phone, or via any agent (including a cloud agent)

Very cool pattern I've been using: 1. Set up a cloud database 2. Set up a boilerplate GitHub repo that can access the cloud database (CLI or small scripts) 3. Put an API key for that cloud database in a 'Claude Code on the web environment' + add the repo Now, you: 1. Can run Claude code on your phone / web and have it load the database with whatever you want 2. Have Claude build artifacts to view the data.. or launch a web app on vercel.. or deploy a mobile app.. 3. Combine it with skills to explore, learn, visualize. I have a full mobile and web app that Claude code can modify on the go. I use Claude code (on my phone!!!) to request new features and it researches, uploads to the database, builds new features in the apps, and pushes.

@DotLasher Awesome!! Glad it's helping you :)

@WillNessAI Yo i tried it, crazily got proper frontend of my application that zooms out for section wise, modifying UI of certain stuffs like text, placements etc, exactly like you showed and my god it made this decisive work more easier! using it with wayfinder now onwards, keep rocking 🗿

Goated thing something similar i was doing but this by far is a crazy thing will for sure gonna try this

@kntl @mattpocockuk @mobbin Have any screenshots ⁉️⁉️

@WillNessAI @mattpocockuk I combined your skill with @mobbin MCP and result is 🤯

I built a variant of @mattpocockuk's grilling skill dedicated to frontend and it has improved how I build new apps and components. The general idea: 1. Use /grilling and /prototype as a base 2. Tell Claude to build 5 WILDLY different prototypes 3. Tell Claude to include a picker that lets you switch between each variant live 4. Each round you select your favorite(s) + leave feedback, and Claude will walk down each branch of the design tree, helping you zoom in on your desired design And THEN, I went and added it to /wayfinder, so whenever I make a new map and there's novel frontend work, a ticket is created specifically referencing that /grilling-frontend-prototyping needs to be invoked. This will not be the last time I build a cool skill and add it to Wayfinder; this is a very powerful pattern for planning work. You can find my skill here: github.com/will-ness-ai/s…

@ajs6888 What do you mean by 'mod the library '?

@WillNessAI 手机上也能改库，这玩法有点爽

@joshsportsbets7 I need to write more about this, but the goal is to have the 'frontend' of this database be a mobile app AND a web app, so I need the database to be in the cloud. Setting up a cloud env in Claude code on web allows my cloud agents to interact with the codebase and database

@WillNessAI /remote control works well for me. Native, and all of the documents/data stay right on your PC where you want them to be. If you set up a cloud then it becomes a different environment then ur PC. Remote control allows me to flip between PC and phone use on the same project

@Gelassoldat - Vercel for web and Expo for mobile. - Mongo for DB

@WillNessAI Lovely idea. What cloud service are you using?

@jaycoolh @mattpocockuk Definitely 😅

@WillNessAI @mattpocockuk Love the idea - skill name needs a rethink though 😂

@Pat_Erichsen @mattpocockuk totally.. tons of opportunities.

@WillNessAI @mattpocockuk this is cool! i feel like this pattern could be applied to so many different domains

@grymaihomelabs @mattpocockuk np! let me know your feedback if you end up using it

@WillNessAI @mattpocockuk wow thats nice! great work and thx for sharing!

@tylermayberry @mattpocockuk Of course! Hope you get good use out of it.

@WillNessAI @mattpocockuk This is great. Thank you for sharing with us.

@mattpocockuk Its been a pretty big unlock for me. At work, I add our custom research agents that have access to internal KBs into wayfinder as well.

@WillNessAI And I love the idea of wayfinder as an orchestrator of other skills, banger