Bruno Hanss

@CesareDadamo Cooodeeeexxxx ! 👋

I want to connect with the people who are into Building with claude/cursor/codex 🛸

@anupamsx hi ! 👋

If you're into tech, let's connect...

@vladsiu @X Hello Vlad, happy to grow together, let's connect !

Hey all - @X I'm looking to #connect with people interested in: - SaaS - Startup - Marketing Let's grow together 🤝

@payaldixit12323 yayyyy you 😀

Hey techies!!! I'm looking to reach to the people interested in: → Frontend → Backend → Full Stack → DevOps → LeetCode → AI/ML → Data Science → UI/UX → Freelancing → Startups Say hi 👋

@nandiny77300 Having fun building prompt-buddy.io, let's connect !

If you're building with AI, let's connect - Founders - vibecoders - Creators

@TanzilaSha9574 Having fun with prompt and skills quality What are you building ?

Startup founders & indie builders 👋 What are you focused on right now? 🚀 🤖 Building AI tools 💻 Launching SaaS products ⚡ Creating automation systems 📱 Developing mobile apps 🌐 Shipping web platforms 🧠 Experimenting with new ideas The best part of 2026? So many people are building and sharing openly 🔥 Drop your startup, app, or side project below 👇 Let’s support and connect with more builders 🤝 #buildinpublic #SaaS #AI #startups #indiehackers

@Akira_Nakamoto_ Hi let's connect

looking to build a network of like minded individuals - creators at the intersection of ai and finance - vibecoders building software with REAL use cases - ai first founders and business owners lets connect

Prompts/instructions can always be improved. I made and use this : prompt-buddy.io/ai-prompt-impr… Let me know what you think if you try it🙏

Here is the prompt to do the skills creation (is used chat gpt 5.5 to get better results) : Goal: Create a persistent AI knowledge system for this monorepo using Codex skills. Inside it, create a main SKILL.md that explains the purpose of the folder, how to navigate the sub-skills, and when each skill should be used. Also configure the repo so that before any non-trivial task, you: 1. Search local skills first. 2. If no suitable local skill exists, use the prompt-buddy MCP server to search for relevant external skills. 3. If an external skill is selected, install it locally into the monorepo. 4. Track what was installed and why. 5. Always preserve brand identity and UI identity when doing non-trivial work. Definitions: A non-trivial task is any task that changes behavior, architecture, design, UI, copy, data flow, tests, build tooling, CI, deployment, package structure, or public API. Only skip this process for simple one-line edits, formatting-only changes, or direct answers that do not modify the repository. Create this structure: .ai/ skills/ SKILL.md repo-practices/ SKILL.md skill-discovery/ SKILL.md brand-identity/ SKILL.md apps/ SKILL.md / SKILL.md ui-identity/ SKILL.md packages/ SKILL.md / SKILL.md Also create or update: AGENTS.md .codex-skills.lock.json Detailed requirements: 1. Main .ai/skills/SKILL.md Create a root skill index with frontmatter: --- name: monorepo-ai-skills description: Use at the start of any non-trivial monorepo task to find the correct local skill for repo practices, brand identity, app implementation details, package implementation details, or UI identity. --- It must explain: - This folder is the canonical AI knowledge base for the monorepo. - Always read this skill first for non-trivial work. - How to choose between repo-practices, brand-identity, skill-discovery, app skills, package skills, and app UI identity skills. - That app/package-specific skills override generic repo practices only for their own scope. - That repo-practices applies everywhere unless a more specific skill says otherwise. - That brand identity must be checked before product, UI, UX, copy, naming, messaging, marketing, onboarding, or visual work. - That UI identity must be checked before modifying an app’s frontend, design system usage, layouts, components, typography, colors, interaction patterns, icons, spacing, animation, empty states, loading states, or error states. - That if no suitable local skill exists, the agent must use prompt-buddy MCP to search for one. 2. .ai/skills/skill-discovery/SKILL.md Create a skill with frontmatter: --- name: skill-discovery description: Use when no suitable local monorepo skill exists for a non-trivial task; searches prompt-buddy MCP, installs selected skills locally, and records them in the skill lockfile. --- It must instruct Codex to: - Inspect .ai/skills first. - Inspect app/package-specific skills if the task touches a known app or package. - Inspect repo-practices for repo-wide rules. - Inspect brand-identity for brand, copy, UI, UX, naming, or messaging work. - If no adequate skill exists, call the prompt-buddy MCP server. - Search by task, language, framework, app name, package name, file paths, error messages, and domain terms. - Prefer narrow, specific skills over broad generic ones. - Save selected external skills under .ai/skills/external//SKILL.md unless a more logical local location exists. - If the external skill is app-specific, save it under .ai/skills/apps///SKILL.md. - If the external skill is package-specific, save it under .ai/skills/packages///SKILL.md. - Update .codex-skills.lock.json. - Record skill name, source, version if available, install path, timestamp, and reason selected. - Read the installed skill before continuing the task. - Do not install duplicate skills. - Do not install skills that conflict with AGENTS.md or repo-practices. 3. .ai/skills/repo-practices/SKILL.md Create a repo-wide skill with frontmatter: --- name: repo-practices description: Use for all non-trivial monorepo work to follow repository-wide conventions that apply across every app and package. --- Populate it by inspecting the repository. Include detailed sections for: - Package manager and install commands. - Monorepo tool, if any, such as Turborepo, Nx, pnpm workspaces, npm workspaces, Yarn workspaces, Lerna, Rush, or custom scripts. - Build commands. - Test commands. - Lint and format commands. - Typecheck commands. - Folder structure. - Naming conventions. - Branching or commit conventions if discoverable. - Environment variable conventions. - Dependency management rules. - Shared code rules. - API/client/server boundaries. - Testing expectations. - Error handling conventions. - Logging conventions. - Security practices. - Performance practices. - Accessibility practices. - Documentation practices. - What not to change without explicit user approval. When unsure, write “Unknown from current inspection” instead of inventing. 4. .ai/skills/brand-identity/SKILL.md Create a brand identity skill with frontmatter: --- name: brand-identity description: Use before changing product copy, UI, UX, naming, messaging, onboarding, marketing pages, visual design, or any user-facing experience in this monorepo. --- Populate it by inspecting the repository. Include detailed sections for: - Product purpose. - Target users. - Brand personality. - Voice and tone. - Naming conventions. - Messaging patterns. - Words and phrases to prefer. - Words and phrases to avoid. - Visual identity. - Color usage if discoverable. - Typography if discoverable. - Iconography if discoverable. - Layout feel. - Motion/animation style if discoverable. - Empty states. - Loading states. - Error states. - Success states. - Accessibility expectations. - Cross-app consistency rules. - Examples from the current implementation. When unsure, write “Unknown from current inspection” instead of inventing. 5. .ai/skills/apps/SKILL.md Create an apps index skill with frontmatter: --- name: monorepo-apps-index description: Use when a task touches one or more applications in the monorepo; routes the agent to the correct app-specific implementation and UI identity skills. --- It must list every app discovered in the repo. For each app, include: - App name. - Path. - Purpose inferred from implementation. - Framework and runtime. - Main entrypoints. - Where routes/pages live. - Where components live. - Where state management lives. - Where API calls live. - Where tests live. - Related packages. - Which skill file to read before touching it. - Which UI identity skill to read before touching its frontend. 6. One app implementation skill per app For every app discovered in the monorepo, create: .ai/skills/apps//SKILL.md Use frontmatter: --- name: app- description: Use before modifying the app, including its routes, components, data flow, state, tests, configuration, build behavior, or runtime behavior. --- Populate with detailed current implementation knowledge: - App purpose. - App path. - Framework. - Runtime. - Build/dev/test commands. - Routing model. - Key entrypoints. - Important directories. - Data fetching strategy. - State management. - Auth/session model if discoverable. - API integration points. - Styling strategy. - Component architecture. - Error handling. - Loading states. - Forms and validation. - Testing setup. - Important dependencies. - Configuration files. - Environment variables. - Known conventions. - Common pitfalls. - Files that should be read before making changes. - How to validate changes. When unsure, write “Unknown from current inspection” instead of inventing. 7. One UI identity skill per app For every app discovered in the monorepo, create: .ai/skills/apps//ui-identity/SKILL.md Use frontmatter: --- name: ui-identity- description: Use before modifying the UI, UX, styling, layout, copy, components, flows, icons, colors, typography, animation, or user-facing states of the app. --- Populate with detailed current UI identity: - UI purpose. - Visual style. - Layout patterns. - Navigation patterns. - Component patterns. - Design system usage. - Color tokens/classes found. - Typography tokens/classes found. - Spacing and radius patterns. - Icon usage. - Animation/motion usage. - Button patterns. - Form patterns. - Table/list/card patterns. - Modal/dialog patterns. - Toast/notification patterns. - Empty states. - Loading states. - Error states. - Success states. - Responsive behavior. - Accessibility patterns. - Copy tone used in the UI. - Screens/routes inspected. - Examples from current components. - Rules for preserving this identity. When unsure, write “Unknown from current inspection” instead of inventing. 8. .ai/skills/packages/SKILL.md Create a packages index skill with frontmatter: --- name: monorepo-packages-index description: Use when a task touches shared packages, libraries, SDKs, configs, tooling packages, UI packages, or reusable modules in the monorepo. --- It must list every package discovered in the repo. For each package, include: - Package name. - Path. - Purpose. - Public API surface. - Main exports. - Build/test/typecheck commands if available. - Dependencies. - Consumers inside the monorepo. - Which skill file to read before touching it. 9. One package implementation skill per package For every package discovered in the monorepo, create: .ai/skills/packages//SKILL.md Use frontmatter: --- name: package- description: Use before modifying the package, including its public API, implementation, tests, build config, exports, dependencies, or consumers. --- Populate with: - Package purpose. - Package path. - Public exports. - Internal structure. - Key files. - Build process. - Test process. - Typecheck process. - Runtime assumptions. - Consumers in the monorepo. - Dependency rules. - API compatibility rules. - Error handling patterns. - Testing patterns. - Common pitfalls. - Files to inspect before making changes. - How to validate changes. When unsure, write “Unknown from current inspection” instead of inventing. 10. Update or create AGENTS.md Add this repo-wide policy: Before any non-trivial task: 1. Read .ai/skills/SKILL.md. 2. Read .ai/skills/repo-practices/SKILL.md. 3. If the task touches an app, read .ai/skills/apps/SKILL.md and that app’s SKILL.md. 4. If the task touches app UI, also read that app’s ui-identity/SKILL.md. 5. If the task touches a package, read .ai/skills/packages/SKILL.md and that package’s SKILL.md. 6. If the task touches brand, copy, UX, UI, naming, or messaging, read .ai/skills/brand-identity/SKILL.md. 7. If no suitable skill exists, use the prompt-buddy MCP server to search for one. 8. If a skill is selected from prompt-buddy, install it under .ai/skills, update .codex-skills.lock.json, read it, and then continue. 9. Do not proceed with non-trivial work until the relevant skills have been read. 10. Preserve the current brand identity and app UI identity unless the user explicitly asks to change them. Also mention: - Do not invent implementation details in skills. - If uncertain, write “Unknown from current inspection”. - Update relevant skills whenever the implementation meaningfully changes. - Keep skills concise enough to be usable, but detailed enough to prevent repeated rediscovery. 11. Create or update .codex-skills.lock.json If missing, create: { "skills": {} } If external skills are later installed, use this shape: { "skills": { "skill-name": { "source": "prompt-buddy", "version": "unknown", "path": ".ai/skills/external/skill-name/SKILL.md", "installed_at": "ISO-8601 timestamp", "reason": "Why this skill was selected" } } } 12. Validation After creating the files: - Print the generated folder tree. - List every app found. - List every package found. - List every SKILL.md created. - Report any areas marked “Unknown from current inspection”. - Do not modify application source code except for creating these AI skill/instruction files.

This turns Codex from: "an agent that edits code" into: "an agent that understands and maintains repo-specific operating memory" That is the difference between one-off automation and consistent/reliable automation.

Codex gets 10x better when it stops rediscovering your repo every task. So i added skills to keep track of my repo rules.