{
  "id": 251564,
  "title": "One skill per action looked like the safe boundary",
  "url": "https://urgent.news/2026/08/07/one-skill-per-action-looked-like-the-safe-boundary",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-08-07T05:24:54.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/michaeltruong/one-skill-per-action-looked-like-the-safe-boundary-13pj"
  },
  "original_language": "en",
  "account": "The article discusses the challenges and solutions encountered while building an AI-assisted editorial pipeline in Cursor. An initial approach of using one skill per action proved to be problematic as the system evolved. The author explains how they initially created an inbox skill that only created cards, but as more work was needed while a card was still in the Inbox, it became clear that this was three distinct jobs. The skills needed to create, enrich, and reclassify cards within the Inbox. However, trying to separate these into three different skills led to issues such as ownership conflicts and unclear routing.\n\nUltimately, the author concluded that consolidating these operations into a single skill was the better solution. The inbox skill now handles create, enrich, and reclassify actions while the card remains in the Inbox. The skill internally routes commands, allowing the user to invoke a single skill and the routing rules deciding the appropriate actions. This consolidation simplifies the system by providing one place owning the Inbox normalization, a shared rule set, and authority over the rules.\n\nThe author suggests applying a \"consolidate-vs-split test\" to similar situations: if the lifecycle stage or artifact type diverges, then the skills should not be consolidated. They also mention that different operations, such as scheduling, drafting, critique, and publishing, own different artifacts and stop lines, so they remain separate. The key takeaway is that a single skill should cover a coherent capability with internal routing between operations, rather than having multiple skills handling related actions.",
  "summary": "I've been building an AI-assisted editorial pipeline in Cursor. Notion cards capture observations, skills score and schedule them, and agents draft markdown that eventually syncs to dev.to. In that setup, a skill is a markdown file the agent loads when you invoke a workflow: it defines what the agent may do, what it must not do, and how it routes between sub-steps. When I started, that model felt…",
  "key_points": [
    "Initial approach of one skill per action proved problematic",
    "Inbox skill needed create, enrich, and reclassify actions",
    "Consolidating operations into a single skill simplifies system"
  ],
  "editors_take": "Consolidating related actions into a single skill with internal routing simplifies the system and provides a coherent capability, resolving issues that arose from separating them into distinct skills.",
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}