{"slug": "agent-skills", "name": "Agent Skills", "category": "ai", "category_name": "AI tools and apps", "description": "", "recommenders": 7, "commenters": 13, "warners": 3, "score": 3.426, "rising": -0.96, "first": "2026-01-15", "latest": "2026-09-07", "why": "I recently switched to using only AGENTS.md and SKILLS.  I appreciate the standardization from this perspective.", "why_is_quote": true, "has_verdict": false, "sentiment": {"recommend": 7, "neutral": 1, "compare": 2, "warn": 3}, "spark": [0, 0, 1, 2, 0, 0, 2, 0, 1, 1, 0, 0], "listed": true, "loved": [], "caveat": "", "short_quote": "I recently switched to using only AGENTS.md and SKILLS.  I appreciate the standardization from this perspective.", "kind": "", "icon": true, "domain": "agentskills.io", "github_repo": null, "official_url": "https://agentskills.io", "pricing": "", "aliases": ["Agent Skills", "agent-skills"], "verdict": null, "praise": [{"label": "", "summary": "", "quotes": [{"id": 789896144, "item_id": 49368509, "url": "https://news.ycombinator.com/item?id=49368509", "by": "mtzaldo", "date": "2026-08-19", "version": null, "stance": "recommend", "quote": "It seems to me he could have use an skill like using-agent-skills from https://github.com/addyosmani/agent-skills go generate the specs", "excerpt": "It seems to me he could have use an skill like using-agent-skills from https://github.com/addyosmani/agent-skills go generate the specs and use a validator like oracle or something along the same lines. Also, a skill like grill-me from Matt P. https://github.com/mattpocock/skills.", "context": "", "use_case": "generating specs for coding agents", "switched_from": "", "story": "Sol Loves to Cheat"}, {"id": 784915648, "item_id": 49057228, "url": "https://news.ycombinator.com/item?id=49057228", "by": "hangrybear666", "date": "2026-07-26", "version": null, "stance": "recommend", "quote": "A good example is the agent skills open standard", "excerpt": "Or other models. I'd much rather have open standards to follow instead of relying on vendor lock in and features automatically enforced by one company, which are not supported by others. A good example is the agent skills open standard which was invented by anthropic but given to the public and is followed by other vendors as well https://agentskills.io/home https://github.com/agentskills/agentskills", "context": "", "use_case": "", "switched_from": "", "story": "The new rules of context engineering for Claude 5 generation models"}, {"id": 768349760, "item_id": 48021860, "url": "https://news.ycombinator.com/item?id=48021860", "by": "kergonath", "date": "2026-05-05", "version": null, "stance": "recommend", "quote": "I\u2019ve had some success with agent skills", "excerpt": "> The slot machine can drop any hard requirement that you specifically in your AGENTS.md, memory.md or your dozens of skill markdowns. Pretty much guaranteed. Indeed. That said, I\u2019ve had some success with agent skills, but I use them to make the LLM aware of things it can do using specific external tools. I think it is a really bad idea to use this mechanism to enforce safety rules. We need good sandboxing for this, ", "context": "", "use_case": "Giving LLM tool awareness", "switched_from": "", "story": "Agent Skills"}, {"id": 768271441, "item_id": 48016965, "url": "https://news.ycombinator.com/item?id=48016965", "by": "AndyNemmity", "date": "2026-05-04", "version": null, "stance": "recommend", "quote": "I commend them for a clean, simple approach.", "excerpt": "This is why I created the /do router, to route to all skills. I also have anti rationalization, progressive context discovery etc. I only make it for me, so it's a bit complex and targeted towards me, and what I do, but it's pretty easy to adjust things. https://github.com/notque/vexjoy-agent Working on reading through Agent Skills, it seems we've converged on a lot of the same points, and I've never seen it, so tryi", "context": "", "use_case": "", "switched_from": "", "story": "Agent Skills"}, {"id": 750050017, "item_id": 46878126, "url": "https://news.ycombinator.com/item?id=46878126", "by": "DonHopkins", "date": "2026-02-03", "version": null, "stance": "recommend", "quote": "Skills work differently. Once loaded into context", "excerpt": "There's a fundamental architectural difference being missed here: MCP operates BETWEEN LLM complete calls, while skills operate DURING them. Every MCP tool call requires a full round-trip \u2014 generation stops, wait for external tool, start a new complete call with the result. N tool calls = N round-trips. Skills work differently. Once loaded into context, the LLM can iterate, recurse, compose, and run multiple agents a", "context": "", "use_case": "Knowledge units for LLMs", "switched_from": "", "story": "Agent Skills"}]}], "caveats": [{"label": "", "summary": "", "quotes": [{"id": 793584321, "item_id": 49599020, "url": "https://news.ycombinator.com/item?id=49599020", "by": "bensyverson", "date": "2026-09-07", "version": null, "stance": "warn", "quote": "it feels like AGENTS.md adherence is higher than Skill adherence", "excerpt": "I'm building something similar with `Agents`[0] (think of it as a package manager for AGENTS.md snippets). I'm getting good results by including detailed instructions for certain tasks in the repo, and providing hints in AGENTS.md about when to use them. It's basically a simpler version of Skills\u2014but it feels like AGENTS.md adherence is higher than Skill adherence. [0]: https://github.com/bensyverson/agents/", "context": "", "use_case": "", "switched_from": "", "story": "Ask HN: How do you manage skills files?"}, {"id": 768261905, "item_id": 48016369, "url": "https://news.ycombinator.com/item?id=48016369", "by": "CharlesW", "date": "2026-05-04", "version": null, "stance": "warn", "quote": "the discoverability of these skills will be difficult without a rename", "excerpt": "From an SEO/LLMO perspective, the discoverability of these skills will be difficult without a rename: https://agentskills.io/ If Addy reads this, how do you pitch this vs. Superpowers? https://github.com/obra/superpowers", "context": "", "use_case": "", "switched_from": "", "story": "Agent Skills"}, {"id": 749944272, "item_id": 46871517, "url": "https://news.ycombinator.com/item?id=46871517", "by": "modernerd", "date": "2026-02-03", "version": null, "stance": "warn", "quote": "In 56% of eval cases, the skill was never invoked.", "excerpt": "That's also what Vercel found: > In 56% of eval cases, the skill was never invoked. The agent had access to the documentation but didn't use it. Adding the skill produced no improvement over baseline. > \u2026 > Skills aren't useless. The AGENTS.md approach provides broad, horizontal improvements to how agents work with Next.js across all tasks. Skills work better for vertical, action-specific workflows that users explici", "context": "", "use_case": "", "switched_from": "", "story": "Agent Skills"}]}], "who": [{"by": "csummers", "context": "developing a new programming language", "url": "https://news.ycombinator.com/item?id=46875734", "date": "2026-02-03"}], "tags": [], "use_cases": ["generating specs for coding agents", "giving llm tool awareness", "knowledge units for llms", "giving llms hints for a new language", "reusable ai agent instructions"], "compared": [{"slug": "agents-md", "name": "AGENTS.md", "co": 3, "switched_from": 0, "switched_to": 0}, {"slug": "superpowers", "name": "Superpowers", "co": 2, "switched_from": 0, "switched_to": 0}], "makers": [{"id": 747095233, "item_id": 46693452, "url": "https://news.ycombinator.com/item?id=46693452", "by": "lixiaofei", "date": "2026-01-20", "version": null, "stance": "neutral", "quote": "", "excerpt": "Recently, \u201cskills\u201d have been blowing up. I collected 60k+ open-source agent skills from GitHub, then curated 1,000+ Claude Code skills that are actually useful. The directory is updated daily. To me, Claude Code skills aren\u2019t that different from agents or other automation tools\u2014the big difference is that they lower the barrier so more people (including beginners) can use them. There are tons of skills in every indust", "context": "", "use_case": "", "switched_from": "", "story": "Show HN: Agent Skills \u2013 1k curated Claude Code skills from 60k+ GitHub skills"}], "threads": [{"url": "https://news.ycombinator.com/item?id=46871173", "title": "Agent Skills", "mentions": 5}, {"url": "https://news.ycombinator.com/item?id=48015397", "title": "Agent Skills", "mentions": 3}, {"url": "https://news.ycombinator.com/item?id=49589914", "title": "Ask HN: How do you manage skills files?", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49348189", "title": "Sol Loves to Cheat", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49051361", "title": "The new rules of context engineering for Claude 5 generation models", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=47712718", "title": "I Still Prefer MCP over Skills", "mentions": 1}], "sentiment_by_month": {"2025-11": {}, "2025-12": {}, "2026-01": {"recommend": 1}, "2026-02": {"warn": 1, "compare": 1, "recommend": 2}, "2026-03": {"neutral": 1}, "2026-04": {"compare": 1}, "2026-05": {"warn": 1, "recommend": 2}, "2026-06": {}, "2026-07": {"recommend": 1}, "2026-08": {"recommend": 1}, "2026-09": {"warn": 1}, "2026-10": {}}, "versions": [], "version_unnamed": 0, "family_ref": null, "runs_on": null, "lines": [], "quotes": [{"id": 793584321, "item_id": 49599020, "url": "https://news.ycombinator.com/item?id=49599020", "by": "bensyverson", "date": "2026-09-07", "version": null, "stance": "warn", "quote": "it feels like AGENTS.md adherence is higher than Skill adherence", "excerpt": "I'm building something similar with `Agents`[0] (think of it as a package manager for AGENTS.md snippets). I'm getting good results by including detailed instructions for certain tasks in the repo, and providing hints in AGENTS.md about when to use them. It's basically a simpler version of Skills\u2014but it feels like AGENTS.md adherence is higher than Skill adherence. [0]: https://github.com/bensyverson/agents/", "context": "", "use_case": "", "switched_from": "", "story": "Ask HN: How do you manage skills files?"}, {"id": 789896144, "item_id": 49368509, "url": "https://news.ycombinator.com/item?id=49368509", "by": "mtzaldo", "date": "2026-08-19", "version": null, "stance": "recommend", "quote": "It seems to me he could have use an skill like using-agent-skills from https://github.com/addyosmani/agent-skills go generate the specs", "excerpt": "It seems to me he could have use an skill like using-agent-skills from https://github.com/addyosmani/agent-skills go generate the specs and use a validator like oracle or something along the same lines. Also, a skill like grill-me from Matt P. https://github.com/mattpocock/skills.", "context": "", "use_case": "generating specs for coding agents", "switched_from": "", "story": "Sol Loves to Cheat"}, {"id": 784915648, "item_id": 49057228, "url": "https://news.ycombinator.com/item?id=49057228", "by": "hangrybear666", "date": "2026-07-26", "version": null, "stance": "recommend", "quote": "A good example is the agent skills open standard", "excerpt": "Or other models. I'd much rather have open standards to follow instead of relying on vendor lock in and features automatically enforced by one company, which are not supported by others. A good example is the agent skills open standard which was invented by anthropic but given to the public and is followed by other vendors as well https://agentskills.io/home https://github.com/agentskills/agentskills", "context": "", "use_case": "", "switched_from": "", "story": "The new rules of context engineering for Claude 5 generation models"}, {"id": 768349760, "item_id": 48021860, "url": "https://news.ycombinator.com/item?id=48021860", "by": "kergonath", "date": "2026-05-05", "version": null, "stance": "recommend", "quote": "I\u2019ve had some success with agent skills", "excerpt": "> The slot machine can drop any hard requirement that you specifically in your AGENTS.md, memory.md or your dozens of skill markdowns. Pretty much guaranteed. Indeed. That said, I\u2019ve had some success with agent skills, but I use them to make the LLM aware of things it can do using specific external tools. I think it is a really bad idea to use this mechanism to enforce safety rules. We need good sandboxing for this, ", "context": "", "use_case": "Giving LLM tool awareness", "switched_from": "", "story": "Agent Skills"}, {"id": 768271441, "item_id": 48016965, "url": "https://news.ycombinator.com/item?id=48016965", "by": "AndyNemmity", "date": "2026-05-04", "version": null, "stance": "recommend", "quote": "I commend them for a clean, simple approach.", "excerpt": "This is why I created the /do router, to route to all skills. I also have anti rationalization, progressive context discovery etc. I only make it for me, so it's a bit complex and targeted towards me, and what I do, but it's pretty easy to adjust things. https://github.com/notque/vexjoy-agent Working on reading through Agent Skills, it seems we've converged on a lot of the same points, and I've never seen it, so tryi", "context": "", "use_case": "", "switched_from": "", "story": "Agent Skills"}, {"id": 768261905, "item_id": 48016369, "url": "https://news.ycombinator.com/item?id=48016369", "by": "CharlesW", "date": "2026-05-04", "version": null, "stance": "warn", "quote": "the discoverability of these skills will be difficult without a rename", "excerpt": "From an SEO/LLMO perspective, the discoverability of these skills will be difficult without a rename: https://agentskills.io/ If Addy reads this, how do you pitch this vs. Superpowers? https://github.com/obra/superpowers", "context": "", "use_case": "", "switched_from": "", "story": "Agent Skills"}, {"id": 763413074, "item_id": 47713317, "url": "https://news.ycombinator.com/item?id=47713317", "by": "bicx", "date": "2026-04-09", "version": null, "stance": "compare", "quote": "I thought about distributing some CLIs with Skills instead.", "excerpt": "I built an internal company MCP that uses Google Workspace auth and injects a mix of guidance (disguised as tools) on how we would like certain tasks to be accomplished via Claude as well as API-like capabilities for querying internal data and safely deploying small apps internally. I\u2019d really love to get away from the SSE MCP endpoints we use, as the Claude desktop app can get really finicky about disconnects. I tho", "context": "", "use_case": "", "switched_from": "", "story": "I Still Prefer MCP over Skills"}, {"id": 757254273, "item_id": 47328392, "url": "https://news.ycombinator.com/item?id=47328392", "by": "mkobit", "date": "2026-03-10", "version": null, "stance": "neutral", "quote": "Agent Skills [2] appear to be the common format across most of the tools", "excerpt": "You could be thinking of the Claude tool search tool [0][1], which lets it do that on-demand loading. Skills (like another comment mentioned) have a progressive disclosure mechanism built-in. Agent Skills [2] appear to be the common format across most of the tools, but I have no idea if this actually has tool provider buy-in. [0]: https://platform.claude.com/docs/en/agents-and-tools/tool-use/tool-search-tool [1]: htt", "context": "", "use_case": "", "switched_from": "", "story": "You Need to Rewrite Your CLI for AI Agents"}, {"id": 750166449, "item_id": 46885403, "url": "https://news.ycombinator.com/item?id=46885403", "by": "DonHopkins", "date": "2026-02-04", "version": null, "stance": "compare", "quote": "(an allegory for Skills -vs- MCP)", "excerpt": "More: Speed of Light -vs- Carrier Pigeon (an allegory for Skills -vs- MCP): https://github.com/SimHacker/moollm/blob/main/designs/SPEED-OF-LIGHT-VS-CARRIER-PIGEON.md", "context": "", "use_case": "", "switched_from": "", "story": "Agent Skills"}, {"id": 750050017, "item_id": 46878126, "url": "https://news.ycombinator.com/item?id=46878126", "by": "DonHopkins", "date": "2026-02-03", "version": null, "stance": "recommend", "quote": "Skills work differently. Once loaded into context", "excerpt": "There's a fundamental architectural difference being missed here: MCP operates BETWEEN LLM complete calls, while skills operate DURING them. Every MCP tool call requires a full round-trip \u2014 generation stops, wait for external tool, start a new complete call with the result. N tool calls = N round-trips. Skills work differently. Once loaded into context, the LLM can iterate, recurse, compose, and run multiple agents a", "context": "", "use_case": "Knowledge units for LLMs", "switched_from": "", "story": "Agent Skills"}, {"id": 750011745, "item_id": 46875734, "url": "https://news.ycombinator.com/item?id=46875734", "by": "csummers", "date": "2026-02-03", "version": null, "stance": "recommend", "quote": "I recently switched to using only AGENTS.md and SKILLS.  I appreciate the standardization from this perspective.", "excerpt": "I'm developing a new programming language, so I have to provide a way for LLMs to know about and generate code for a language they have not seen (i.e., have no training data for). My tooling was previously adding in AI hints with CLAUDE.md, Cursor Rules, Windsurf Rules, AGENTS.md, etc., but I recently switched to using only AGENTS.md and SKILLS. I appreciate the standardization from this perspective.", "context": "developing a new programming language", "use_case": "giving LLMs hints for a new language", "switched_from": "", "story": "Agent Skills"}, {"id": 749966128, "item_id": 46872883, "url": "https://news.ycombinator.com/item?id=46872883", "by": "postalcoder", "date": "2026-02-03", "version": null, "stance": "compare", "quote": "Vercel ran a study where they said that skills underperform putting a docs index in AGENTS.md", "excerpt": "Your skepticism is valid. Vercel ran a study where they said that skills underperform putting a docs index in AGENTS.md[0]. My guess is that the standardization is going to make its way into how the models are trained and Skills are eventually going to pull out ahead. 0: https://vercel.com/blog/agents-md-outperforms-skills-in-our-agent-evals", "context": "", "use_case": "", "switched_from": "", "story": "Agent Skills"}, {"id": 749944272, "item_id": 46871517, "url": "https://news.ycombinator.com/item?id=46871517", "by": "modernerd", "date": "2026-02-03", "version": null, "stance": "warn", "quote": "In 56% of eval cases, the skill was never invoked.", "excerpt": "That's also what Vercel found: > In 56% of eval cases, the skill was never invoked. The agent had access to the documentation but didn't use it. Adding the skill produced no improvement over baseline. > \u2026 > Skills aren't useless. The AGENTS.md approach provides broad, horizontal improvements to how agents work with Next.js across all tasks. Skills work better for vertical, action-specific workflows that users explici", "context": "", "use_case": "", "switched_from": "", "story": "Agent Skills"}, {"id": 746296736, "item_id": 46643546, "url": "https://news.ycombinator.com/item?id=46643546", "by": "jameslk", "date": "2026-01-15", "version": null, "stance": "recommend", "quote": "they're certainly superior to MCP in ways, being much easier to implement", "excerpt": "> 3 months ago, Anthropic and Simon claimed that Skills were the next big thing and going to completely change the game. So far, from my exploration, I don't see any good examples out there, nor is a there a big growing/active community of users. Skills have become widely adopted since Anthropic's announcement. They've been implemented across major coding agents[0][1][2] and standardized as a spec[3]. I'm not sure wh", "context": "", "use_case": "Reusable AI agent instructions", "switched_from": "", "story": "First impressions of Claude Cowork, Anthropic's general agent"}]}