{"slug": "agents-are-learning-to-write-their-own-skill-md-files", "title": "Agents Are Learning to Write Their Own SKILL.md Files", "summary": "A developer built a demo showing AI agents that can write their own SKILL.md files, enabling them to reuse learned procedures and cut total effort nearly in half. The Agent Skills open standard, introduced by Anthropic in October 2025 and published in December 2025, allows skills to be shared across 30+ agent tools. The demo follows the standard exactly, producing skills that load in stages to minimize context usage.", "body_md": "*The Agent Skills open standard today, and the 2026 research on agents that write their own skills.*\n\n**TL;DR:** In late 2025, \"Agent Skills\" became a thing — a dead-simple way to teach an AI agent a task: a folder with a `SKILL.md`\n\nfile (some instructions in Markdown). It's already an open standard. The wild part is what's coming next: agents that **write their own skills.** I built a demo where an agent solves a task the hard way once, saves a real `SKILL.md`\n\n, and then reuses it — cutting its total effort almost in half. ~130 lines, no API key.\n\nIf you've used Claude Code or similar tools lately, you've probably seen `SKILL.md`\n\nfiles. The idea is refreshingly low-tech. A \"skill\" is just a folder with a Markdown file that says *how to do something*:\n\n```\n---\nname: csv-to-markdown\ndescription: Turn comma-separated text into a Markdown table. Use when the input looks\n  like CSV and the user wants a table.\n---\n\n# CSV to Markdown\n\n## Instructions\nSplit the text into rows on newlines and columns on commas. Make the first row the\nheader, add a `---` divider row, then format every row as `| a | b | c |`.\n```\n\nThat's it. No SDK, no config. Anthropic introduced this in October 2025 and then published it as an **open standard** ([agentskills.io](https://agentskills.io)) in December 2025, so the same skill folder now works across ~30+ different agent tools (Claude Code, Cursor, Copilot, and more).\n\nThe full rules are short ([agentskills.io/specification](https://agentskills.io/specification)): the only **required** fields are `name`\n\n(1–64 chars, lowercase-with-hyphens, and it must match the folder name) and `description`\n\n(≤1024 chars, saying *what it does and when to use it*). Everything else — `license`\n\n, `metadata`\n\n, `compatibility`\n\n, `allowed-tools`\n\n— is optional. That's the whole spec. The `SKILL.md`\n\nfiles my demo writes follow it to the letter, so they'd load unmodified in any compatible CLI.\n\nHere's the smart part. If you just dumped 50 skills' worth of instructions into the agent's context, you'd fill it up and leave no room for actual work. So skills load in **stages**:\n\n`name`\n\nand one-line `description`\n\nof every skill (tiny).So the agent can have *hundreds* of skills installed and barely pay for it — it only reads the short descriptions until one matches, then pulls in the details. My demo shows the math: to use 1 skill out of 3 installed, loading everything costs ~1500 \"tokens\"; the SKILL.md way costs ~560. That gap gets huge as your library grows.\n\nThis is also why people say skills and **MCP** are teammates, not rivals: MCP is how an agent *connects to tools*; a skill is how an agent *knows the procedure* for using them.\n\nToday, humans write `SKILL.md`\n\nfiles. The 2026 research is about agents that write their **own** — and get better over time as their skill library grows. This goes back to **Voyager** (2023), an agent that played Minecraft and saved working code as reusable skills, getting dramatically faster at the game. The new wave makes it general:\n\n`.memory.md`\n\nof notes about itself.The common thread: **solve it once, save the recipe, reuse it forever** — and let the collection get smarter on its own.\n\n📄\n\nThe \"this is the future\" link:Anthropic's own writeup,[, and the open standard at]Equipping agents for the real world with Agent Skills. For the research direction,[agentskills.io][MUSE-Autoskill (arXiv:2605.27366)]and[Skill-Pro (arXiv:2602.01869)]are the clearest reads on agents that grow their own skill libraries.\n\nThis isn't theoretical — the exact pattern from my demo already ships in coding CLIs. In **Claude Code**, a skill is just a folder under `.claude/skills/`\n\nin your repo:\n\n```\n# Anywhere in your project — drop a skill in and the CLI auto-discovers it\nmkdir -p .claude/skills/csv-to-markdown\n$EDITOR .claude/skills/csv-to-markdown/SKILL.md   # same SKILL.md format as my demo\n```\n\nNow the agent loads only that skill's one-line `description`\n\nuntil a task matches — then pulls in the full instructions (that's progressive disclosure doing its job). Type `/skills`\n\ninside the CLI to see what's loaded.\n\nThe best part: because it's an **open standard**, the *same* folder works unmodified across tools. You're not locked in:\n\nWrite the skill once, use it everywhere. The future bit my demo points at: instead of *you* hand-writing that file, the agent writes it for itself after solving the task the first time — and from then on, your repo quietly accumulates a library of skills your agent earned.\n\nSame stream of 7 tasks. \"Cost\" is how much effort each one took.\n\n| No-skills agent | Skill-writing agent | |\n|---|---|---|\n| What it does | re-solves everything from scratch | learns a task once, saves a `SKILL.md` , reuses it |\n| Total cost | 35 |\n19 |\n| Both correct? | 7/7 | 7/7 |\n\n```\n[5] csv-to-markdown  learned it and wrote SKILL.md\n[5] slugify          learned it and wrote SKILL.md\n[1] csv-to-markdown  reused skill 'csv-to-markdown'   ← cheap now\n[5] extract-emails   learned it and wrote SKILL.md\n[1] slugify          reused skill 'slugify'\n[1] csv-to-markdown  reused skill 'csv-to-markdown'\n[1] extract-emails   reused skill 'extract-emails'\n```\n\nIt writes **real SKILL.md files** into a\n\n`./skills`\n\nfolder you can open. The first time it sees a task it pays full price; after that, it finds its own saved skill and reuses it for cheap.Two big reasons engineers should care:\n\n**Agents stop repeating themselves.** Right now most agents re-derive the same thing over and over, paying for it every time. A skill library means \"figure it out once, then it's free\" — like a teammate who writes things down instead of relearning them daily.\n\n**A whole new ecosystem.** There are already 65,000+ shared skills and a scramble to build \"the npm of agent skills\" — registries and marketplaces where you install a skill like a package. Skills are becoming a unit of *shareable expertise*: a senior engineer's know-how, packaged in a folder, that any agent can pick up.\n\nTools tell an agent\n\nwhat it can do. Skills tell ithow to do things well— and soon, agents will write that part themselves, and trade it with each other.\n\n```\ngit clone https://github.com/Shridhar-2205/living-software\ncd living-software/06-agent-skills\npython demo.py\ncat skills/csv-to-markdown/SKILL.md   # a skill the agent wrote itself\n```\n\nHonest note: this is a POC. Real systems decide *when* a new skill is worth saving, test it, and refine it over time (that's exactly what the 2026 papers above tackle). Mine keeps that part simple so the core idea — *learn once, save a SKILL.md, reuse it* — is easy to see.\n\n*Written by **Shridhar Shah**, Senior Software Engineer at Outshift by Cisco — I work on AI agents, search, and how they \"think.\" Part 6 of \"Toward Living Software.\" GitHub · LinkedIn*\n\nSources:Anthropic, \"Equipping agents for the real world with Agent Skills\" (2025) and the Agent Skills open standard (agentskills.io); Voyager (arXiv:2305.16291); MUSE-Autoskill (arXiv:2605.27366); Memento-Skills (arXiv:2603.18743); Skill-Pro (arXiv:2602.01869); MemSkill (arXiv:2602.02474).", "url": "https://wpnews.pro/news/agents-are-learning-to-write-their-own-skill-md-files", "canonical_source": "https://dev.to/shridhar_shah2297/agents-are-learning-to-write-their-own-skillmd-files-3foo", "published_at": "2026-06-27 21:53:22+00:00", "updated_at": "2026-06-27 22:03:31.526147+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-agents", "developer-tools"], "entities": ["Anthropic", "Claude Code", "Cursor", "Copilot", "Voyager", "MUSE-Autoskill", "Skill-Pro", "agentskills.io"], "alternates": {"html": "https://wpnews.pro/news/agents-are-learning-to-write-their-own-skill-md-files", "markdown": "https://wpnews.pro/news/agents-are-learning-to-write-their-own-skill-md-files.md", "text": "https://wpnews.pro/news/agents-are-learning-to-write-their-own-skill-md-files.txt", "jsonld": "https://wpnews.pro/news/agents-are-learning-to-write-their-own-skill-md-files.jsonld"}}