Agentic Coding
Top tech stories today across software, hardware, AI, and product launches. Senior engineer audience — skip rumour churn and pre-announcement leaks. Lead with shipping-now stories and what changed for builders.
Supabase ships plug-and-play skills, oh-my-openagent adds multi-agent orchestration
1 Min. Lesezeit
Supabase agent skills
Supabase just handed you a shortcut for database work.
Their new agent-skills repository bundles reusable skill modules for Claude Code, Cursor, and Cline—covering databases, auth, edge functions, and PostgreSQL best practices out of the box [Source: GitHub]. Install via command line or as Claude Code plugins, and the skills auto-trigger when your agent detects relevant tasks like query optimization or schema reviews. The open standard format means you can inspect manifests and reference docs before enabling anything.
Add these to your curated skill stack from yesterday's update—just mind your budget fraction.
oh-my-openagent orchestration
Multi-agent orchestration finally has a turnkey install.
oh-my-openagent ships with Sisyphus as your main orchestrator, delegating deep work to Hephaestus and strategic planning to Prometheus [Source: GitHub]. It auto-injects AGENTS.md for project context, uses hash-anchored edits for precision, and supports Team Mode for parallel multi-agent execution. The /init-deep command generates hierarchical AGENTS.md files automatically—no manual setup required.
If you liked Ruflo's parallel swarms, this gives you the same firepower with a planning layer baked in.
Anthropic's recursive improvement
Anthropic just published how Claude builds itself.
Their new research paper details metrics on AI-assisted development inside Anthropic—not a how-to guide for your workflow, but a window into where agentic coding is heading [Source: Anthropic]. The focus is internal benchmarks and future scenarios for recursive self-improvement. Worth reading for context on why context management and skill budgets matter more as models get better at modifying their own code.
File this under strategic awareness rather than weekend implementation.
Agent Skills to help developers using AI agents with Supabase16 hours ago ... ... agent-skills --skill supabase-postgres-best-practices. Claude Code Plugin. You can also install the skills as Claude Code plugins: # 1. Install supabase/agent ...github.com
The Supabase Agent Skills repository provides a collection of reusable skill modules designed to enhance AI agents like Claude Code, Cursor, and Cline with structured instructions and resources. The available skills include comprehensive Supabase development guidance covering databases, authentication, edge functions, and integrations, plus PostgreSQL best practices optimized across critical categories like query performance, connection management, and security. These skills can be installed via command line or as Claude Code plugins, and are automatically discovered and used by agents when relevant tasks are detected—such as optimizing queries, reviewing schemas, setting up authentication, or implementing indexes. The skills follow an open standard format with manifests and reference documentation, enabling developers to work more efficiently with AI agents on structured, best-practice-guided tasks.
GitHub - code-yeongyu/oh-my-openagent: omo/lazycodex10 hours ago ... An LLM agent reads the full guide and walks every step correctly. Paste this prompt into Claude Code, AmpCode, Cursor, or any agent: Install and configure oh-my ...github.com
The user's intent is clearly in English (title and description are in English), so I'll provide the summary in English. The website content describes oh-my-openagent, an AI agent orchestration framework for code development. Relevant to the user's interests in agentic coding, PRD workflows, and advanced AI pair programming: oh-my-openagent provides multi-agent orchestration with Sisyphus as the main orchestrator delegating to specialist agents (Hephaestus for autonomous deep work, Prometheus for strategic planning). It includes AGENTS.md auto-injection for project-specific context, hash-anchored edit tools for precise code modifications, and Team Mode for parallel multi-agent execution. The framework supports both OpenCode (Ultimate Edition) and Codex CLI (Light Edition) with the ultrawork command enabling autonomous task completion without manual configuration. Key features include Skills with embedded MCP servers, Prometheus interview-mode planning before execution, /init-deep for hierarchical AGENTS.md generation, and LSP integration for IDE-precision agent workflows. Installation is available via bunx oh-my-openagent install or npx lazycodex-ai install, with configuration through AGENTS.md and rules injection.
When AI builds itself - Anthropic1 hour ago ... How to read this: Session success is determined by a Claude judge; a session is deemed successful if the Claude Code agent clearly succeeded at the user's tasks ...anthropic.com

I found no relevant news in this content for your stated interests. The website content is a detailed research paper from Anthropic about recursive self-improvement and AI-assisted development at their organization. While it discusses Claude's capabilities in coding and research workflows, it does not cover the specific topics you're seeking: Claude Code AGENTS.md setup procedures, agentic coding PRD workflow patterns, or Cursor AI pair programming techniques. The content focuses on internal metrics and future AI development scenarios rather than practical implementation guides, best practices documentation, or YouTube tutorials for advanced "vibe coding" with these tools.