Tag
A Twitter thread clarifies the distinct roles of Skills, Subagents, MCP, and Hooks in AI agent design, arguing that confusing them leads to inefficient and costly systems.
Google Antigravity has introduced support for building autonomous teams of specialized agents, including roles like Sentinel, Orchestrator, Explorer, Worker, Reviewer, Critic, and Auditor, each with distinct responsibilities rather than a single generalist agent.
Developer @realWeZZard shared his experience of building the charge plugin on his own four months before Anthropic launched Dynamic Workflow, detailing the design trade-offs of subagent orchestration, cost control, and the reasons for ultimately choosing amplify, and provided a practical decision tree.
The article analyzes a 2026 paper by Bai et al. showing that subagents and context bloat cause token costs in long agent runs to be ~1000x higher than chat, and presents three practical fixes (PLAN.md, read budget, out-of-band notes) that reduce token usage by 70-90%.
Google AI Devs showcases how parallel subagents in Antigravity can automatically sort and rename hundreds of marketing assets, reducing manual file management.
Codex introduced new capabilities for creating and managing threads across projects and tasks, including orchestration of worktrees and use of subagents for parallel tasks.
Claude Code introduces a dynamic workflow feature where Claude writes orchestration scripts, spawns subagent swarms, and verifies results, accessible via /model opus 4.8 and /effort ultracode.
Claude Code introduces dynamic workflow capabilities, generating orchestration scripts in real time and running a large number of sub-agents in parallel to handle complex tasks. Supports Max, Team, Enterprise, and API endpoints, but token consumption is significantly higher than normal sessions.
Claude Code introduces dynamic workflows, allowing Claude to write orchestration scripts and spawn coordinated subagents for complex tasks.
Claude Code's research preview now includes dynamic workflows, allowing Claude to write orchestration scripts and spin up coordinated subagents for complex tasks.
A tweet highlights that Hermes Agent can identify repeated workflows and turn them into reusable skills or custom subagents, with a call to action for Codex users.
A tweet introduces a workflow where GPT-5.5 xhigh plans and delegates implementation to Composer 2.5 subagents via the pi-cursor-sdk, claiming it outperforms using either model alone. The linked GitHub repo is an open-source SDK that integrates Cursor models into the pi agent runtime.
Introduces an Awesome repository containing over 100 sub-agents that upgrade Claude Code from a single-agent assistant to a multi-agent collaborative AI development platform, and provides a quick-start guide and top sub-agent recommendations.
A tip for using Codex to turn repeated prompts into reusable skills and subagents, improving workflow efficiency for tasks like CI failure checking and PR review.
A detailed guide on setting up custom subagents for Codex, using a grid of six generic agents with varying effort and permission levels, plus a mission card pattern for efficient task delegation.
A framework for structuring Claude Code as an AI operating system using CLAUDE.md, skills, hooks, subagents, and plugins to create scalable, automated workflows.
This thread shares strategies to reduce token usage in AI agents, including prompt caching, context summarization, using smaller models, trimming tool outputs, subagents, RAG, and tight system prompts.
A researcher shares an observation in evaluating subagent behavior within deep agent systems, noting an interesting quirk in how agents align with hand-written system prompts versus instructions from the orchestrator.
Demonstrates running subagents locally on a MacBook Pro M5 using Codex CLI and LM Studio with Qwen 3.6 and MLX batching for code review and bug detection.
Claude Code's plugin system allows powerful extensions beyond skills, but its proprietary nature risks vendor lock-in as only a few competitors support it.