#agentskills — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #agentskills, aggregated by home.social.
-
I looked at 1,800 entries from skills.sh to see what people publish as agent skills.
Coding is important, but it is not most of the trending or hot lists. Many skills cover workplace automation, content, marketing, and the way people prefer to work.
That matters if your company wants an internal marketplace. A large catalog is easy. Keeping local instructions owned and current takes more work.
https://www.the-main-thread.com/p/agent-skills-company-marketplaces
-
#OpenAI is celebrating GPT-5’s first anniversary by introducing #AgentPlugins, an open standard for reusable #AIagent #extensions. This vendor-neutral standard, developed in collaboration with companies like Amazon and Microsoft, aims to create a consistent format for #AgentSkills and #MCPservers. While OpenAI hasn’t announced GPT-6, the unreleased Astra model family shows promise, having made advancements in longstanding problems. https://9to5mac.com/2026/08/06/gpt-5-turning-one-as-openai-shares-new-agent-plugins-standard/?eicker.news #tech #news #ainews
-
"Consider this idea: A company has 100 tech-writer-specific skills that are highly adopted and used across the organization. This collection of skills, focusing on internal authoring skills, has skills for doing virtually anything related to tech writing that you can imagine — fixing a bug, applying a style edit, checking release notes for accuracy, fixing comments in proto styles, and more. The skills don’t do the job entirely themselves but rather act as power tools for the writers, accelerating and amplifying their work.
Not everyone uses all 100 of the tech writer skills. Instead, each tech writer registers the skills most relevant to their tasks in their agent’s configuration. Some have even created virtual agents that have these skills.
Is this 100 TW skills idea worthwhile? The merits of the 100 skills idea are questionable; there are many problems associated with it — skills are hard to share because tech writing processes are idiosyncratic, monolithic skills that try to do too much get rejected, and there’s a fundamental trust problem when running someone else’s skill on your content. And yet, I think the idea could be interesting. We would essentially be externalizing the skills of an entire profession in a way that could be used cross-organizationally, by any role. This is unsettling as it may lead to our irrelevance. But there’s also a high possibility of amplification: by externalizing our skills, we build upon each others’ strengths. Is it possible to construct skills in such a way that other writers, particularly those who didn’t make the skills, find them useful?
In this series, I’ll build out these ideas into a course on skills."
https://idratherbewriting.com/ai/skills.html
#TechnicalWriting #SoftwareDocumentation #AI #AIAgents #GenerativeAI #LLMs #AgenticAI #AgentSkills
-
🎯 #TasteSkill is a set of portable #AgentSkills that give #AI coding agents design taste: stronger layout, typography, motion and spacing instead of boilerplate-looking UIs #opensource #frontend
🧵👇 -
Key inspirations behind ORF:
• Google’s #ReasoningBank paper (traps & resolution paths)
• Open Knowledge Format (OKF)
• @karpathy’s #LLMWiki concept
• Progressive disclosure in #AgentSkillsI also brainstormed much of the core concept with Gemini! 🤖
-
Agent skills start as Markdown, then quietly become part of the system.
This walkthrough uses skillsaw with IBM Bob: build a small invoice-processing skill, add a reference file, lint it, pin the rule set, wire CI, and test the behavior locally.
No MCP server. No API keys. Just a SKILL.md that is reviewable.
-
Everyone gives agents skills.
I made skills hatch their own agents.SKILL.md → 6-Harness Compiler → Standalone Python Agent.
Not a prompt wrapper. Actual generated code.
3 commands from markdown to a running agent.No host agent required. No context leakage.
15KB of prompts → 150 bytes of runtime config. -
quicktipp #116: In a Claude Skill document you can dynamically run shell commands and inject their output into the skill document before it is sent to the model. Pipes, grepping, etc all works.
The syntax is:
!`your-shell-command`#claude #claudecode #agents #agentskills #skills #llm #ai #vibecoding
-
Agent skills are not prompt snippets with a nicer file name.
I wrote about the part that matters for real developer work: skills as workflow contracts, plus hooks, scripts, approval points, and QA gates that make repeated agent tasks safer to run.
-
#Development #Debates
The great agent skills land grab · When do agent skills actually matter? https://ilo.im/16dnau_____
#Programming #Coding #AI #AiAgents #AgentSkills #SkillMd #ClaudeCode #WebDev #Frontend #Backend
