[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7DEF4ZHQB8AYQYGDXS2V032":3,"skill-same:01M7DEF4ZHQB8AYQYGDXS2V032":44},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":14,"repositoryHtmlUrl":27,"repositoryHomepageUrl":6,"ownerHtmlUrl":28,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":29,"install":30,"resources":43},"01M7DEF4ZHQB8AYQYGDXS2V032","skill-authoring",null,"Write, adapt or vendor an Agent Skill for this repository (a SKILL.md directory under .agents\u002Fskills\u002F). Use when asked to create a skill, turn a procedure into one, bring in a skill from elsewhere, or review one against the repository's skill convention.","romeroabelleira\u002Fxeito","romeroabelleira","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F108562?v=4","01M7DEES0DQ09Z5M83W75B3J7C","Juan Romero Abelleira","01M7DEES13Q7TFAGMGG4P99S49",0,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fromeroabelleira\u002Fxeito\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fskill-authoring\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: skill-authoring\ndescription: Write, adapt or vendor an Agent Skill for this repository (a SKILL.md directory under .agents\u002Fskills\u002F). Use when asked to create a skill, turn a procedure into one, bring in a skill from elsewhere, or review one against the repository's skill convention.\n---\n\n# Skill authoring\n\nSkills here follow the convention in `.agents\u002Fskills\u002FREADME.md`. Read it first; this skill is how to apply it.\n\n## Where and how\n\n- One directory per skill: `.agents\u002Fskills\u002F\u003Cname>\u002FSKILL.md`. The name is lowercase words joined by hyphens, and the frontmatter's `name` is exactly the directory name.\n- Long material goes into `references\u002F\u003Ctopic>.md`, linked from `SKILL.md` with a line saying when to read it. Code the skill runs goes into `scripts\u002F`, templates into `assets\u002F`.\n\n## The description decides whether the skill is used\n\nThe description is the only part an agent always sees, next to every other skill's (xeito keeps the first 1,024 characters). An agent decides from it alone whether to load the skill, so it has to say:\n- **what** the skill does, in the words a request would use;\n- **when** to use it: the situations and phrasings that should trigger it, and, if a neighbouring skill covers part of the ground, where the line is (\"use otp for process design\").\n\nToo narrow and the skill is never loaded; too broad and it crowds out better ones.\n\nXeito does not show the user's skills to every turn: it shortlists them by the request's words (`Xeito.Skills.Index`). Words a request would use that the description cannot hold go under `metadata: keywords:` (`metadata:` then an indented `keywords: word, word`). One keyword is enough to shortlist the skill, so keep each specific, and each part of a hyphenated one too (\"stress-test\" matches every request about a test).\n\n## The body\n\n- Write instructions in the imperative, and say **why** where the reason is not obvious: an agent that knows the reason handles the case the instruction did not foresee. Capital-letter rules without reasons are brittle.\n- Every line should change what the agent does. Leave out what any capable agent already knows.\n- Keep it short; a `SKILL.md` past a few hundred lines belongs split into `references\u002F`.\n- Write for any agent (convention point 2): describe the action, not a particular tool. Use the project's commands, and the project's check is `mix ci`.\n- Nothing about a particular deployment: no hosts, personal paths, hardware or model tags.\n\n## Bringing in a skill from elsewhere\n\n1. Read all of it first, scripts included: a skill is instructions an agent will follow, so check what it would make an agent do.\n2. Check the license allows copying and changing, and copy it as `LICENSE`.\n3. Copy the skill unchanged into `.agents\u002Fskills\u002F\u003Cname>\u002F`.\n4. Change only what conflicts with the convention or the project: tool-specific commands, its own checks or hooks, pointers to skills that do not exist here.\n5. Write `ATTRIBUTION.md`: source, path, upstream commit (full hash), date, and `## Changes` with each change and its reason. Add the one-line notice at the top of a changed `SKILL.md`.\n6. If most of it would have to change, write a skill of your own instead.\n\n## Test it against not having it\n\nA skill earns its place only if agents do better with it than without it, and only if it is loaded when it should be. Check both before adding one, and again after a large change:\n\n1. Write down 3–5 realistic requests the skill is for, and 1–2 near misses that a neighbouring skill or no skill should handle. For each request, write what a good result must contain *before* running anything, so the result cannot bend the expectation.\n2. Run every request twice, each in a fresh session: once with the skill, once without it (a scratch copy of the workspace without the skill's directory). Compare both results against the expectations.\n3. Check triggering with plain requests, not forced ones: the skill should be loaded for the requests it is for and not for the near misses. A miss in either direction is a description problem.\n4. If the session without the skill does as well, drop the skill, or cut it down to the part that made the difference.\n\nKeep the requests and expectations with the change that adds the skill (in the commit message or the plan), so the next change to the skill can rerun them. The method follows the evaluation loop of Anthropic's `skill-creator` skill (Apache-2.0), without its scripts.\n\n## Before you finish\n\n- `mix test test\u002Fxeito\u002Frepo_skills_test.exs` checks loading, naming, the description, tool-specific commands and attribution. `mix ci` runs it too.\n- Rerun the requests from the test above. In xeito, `\u002Fskill:\u003Cname> \u003Crequest>` forces a skill; the transcript shows when a plain request loaded it.\n- Add the skill to the table in `.agents\u002Fskills\u002FREADME.md` and to the list in `AGENTS.md`.\n","# Skill authoring\n\nSkills here follow the convention in `.agents\u002Fskills\u002FREADME.md`. Read it first; this skill is how to apply it.\n\n## Where and how\n\n- One directory per skill: `.agents\u002Fskills\u002F\u003Cname>\u002FSKILL.md`. The name is lowercase words joined by hyphens, and the frontmatter's `name` is exactly the directory name.\n- Long material goes into `references\u002F\u003Ctopic>.md`, linked from `SKILL.md` with a line saying when to read it. Code the skill runs goes into `scripts\u002F`, templates into `assets\u002F`.\n\n## The description decides whether the skill is used\n\nThe description is the only part an agent always sees, next to every other skill's (xeito keeps the first 1,024 characters). An agent decides from it alone whether to load the skill, so it has to say:\n- **what** the skill does, in the words a request would use;\n- **when** to use it: the situations and phrasings that should trigger it, and, if a neighbouring skill covers part of the ground, where the line is (\"use otp for process design\").\n\nToo narrow and the skill is never loaded; too broad and it crowds out better ones.\n\nXeito does not show the user's skills to every turn: it shortlists them by the request's words (`Xeito.Skills.Index`). Words a request would use that the description cannot hold go under `metadata: keywords:` (`metadata:` then an indented `keywords: word, word`). One keyword is enough to shortlist the skill, so keep each specific, and each part of a hyphenated one too (\"stress-test\" matches every request about a test).\n\n## The body\n\n- Write instructions in the imperative, and say **why** where the reason is not obvious: an agent that knows the reason handles the case the instruction did not foresee. Capital-letter rules without reasons are brittle.\n- Every line should change what the agent does. Leave out what any capable agent already knows.\n- Keep it short; a `SKILL.md` past a few hundred lines belongs split into `references\u002F`.\n- Write for any agent (convention point 2): describe the action, not a particular tool. Use the project's commands, and the project's check is `mix ci`.\n- Nothing about a particular deployment: no hosts, personal paths, hardware or model tags.\n\n## Bringing in a skill from elsewhere\n\n1. Read all of it first, scripts included: a skill is instructions an agent will follow, so check what it would make an agent do.\n2. Check the license allows copying and changing, and copy it as `LICENSE`.\n3. Copy the skill unchanged into `.agents\u002Fskills\u002F\u003Cname>\u002F`.\n4. Change only what conflicts with the convention or the project: tool-specific commands, its own checks or hooks, pointers to skills that do not exist here.\n5. Write `ATTRIBUTION.md`: source, path, upstream commit (full hash), date, and `## Changes` with each change and its reason. Add the one-line notice at the top of a changed `SKILL.md`.\n6. If most of it would have to change, write a skill of your own instead.\n\n## Test it against not having it\n\nA skill earns its place only if agents do better with it than without it, and only if it is loaded when it should be. Check both before adding one, and again after a large change:\n\n1. Write down 3–5 realistic requests the skill is for, and 1–2 near misses that a neighbouring skill or no skill should handle. For each request, write what a good result must contain *before* running anything, so the result cannot bend the expectation.\n2. Run every request twice, each in a fresh session: once with the skill, once without it (a scratch copy of the workspace without the skill's directory). Compare both results against the expectations.\n3. Check triggering with plain requests, not forced ones: the skill should be loaded for the requests it is for and not for the near misses. A miss in either direction is a description problem.\n4. If the session without the skill does as well, drop the skill, or cut it down to the part that made the difference.\n\nKeep the requests and expectations with the change that adds the skill (in the commit message or the plan), so the next change to the skill can rerun them. The method follows the evaluation loop of Anthropic's `skill-creator` skill (Apache-2.0), without its scripts.\n\n## Before you finish\n\n- `mix test test\u002Fxeito\u002Frepo_skills_test.exs` checks loading, naming, the description, tool-specific commands and attribution. `mix ci` runs it too.\n- Rerun the requests from the test above. In xeito, `\u002Fskill:\u003Cname> \u003Crequest>` forces a skill; the transcript shows when a plain request loaded it.\n- Add the skill to the table in `.agents\u002Fskills\u002FREADME.md` and to the list in `AGENTS.md`.\n",{"name":5,"description":7},"main","a13c65918461608e1ea5b3ecb204ef8cf06964b0",".agents\u002Fskills\u002Fskill-authoring","https:\u002F\u002Fgithub.com\u002Fromeroabelleira\u002Fxeito","https:\u002F\u002Fgithub.com\u002Fromeroabelleira","2026-11-13 03:05:42",{"source":8,"skillName":5,"commands":31,"prompt":41,"zipAvailable":42},[32,35,38],{"tool":33,"command":34},"npx","npx skills add romeroabelleira\u002Fxeito --skill skill-authoring",{"tool":36,"command":37},"bunx","bunx skills add romeroabelleira\u002Fxeito --skill skill-authoring",{"tool":39,"command":40},"pnpm","pnpm dlx skills add romeroabelleira\u002Fxeito --skill skill-authoring","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fromeroabelleira\u002Fxeito\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fskill-authoring\u002FSKILL.md",true,[],[45,52,59,66,73,80],{"skillNo":46,"name":47,"nameZh":6,"description":48,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":49,"categories":50,"keywords":51,"publishedAt":19},"01M7DEG5Y3ZSHKHAG3T4KF4A86","web-search","Search the web and extract the content of the result pages.","https:\u002F\u002Fgithub.com\u002Fromeroabelleira\u002Fxeito\u002Fblob\u002Fmain\u002Fpriv\u002Fskills\u002Fbench\u002Fskills\u002Fweb-search\u002FSKILL.md",[],[],{"skillNo":53,"name":54,"nameZh":6,"description":55,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":56,"categories":57,"keywords":58,"publishedAt":19},"01M7DEG3GVYEGJ1RJP7P4N90MF","video-transcript","Fetch the transcript of a YouTube video for summarization and analysis.","https:\u002F\u002Fgithub.com\u002Fromeroabelleira\u002Fxeito\u002Fblob\u002Fmain\u002Fpriv\u002Fskills\u002Fbench\u002Fskills\u002Fvideo-transcript\u002FSKILL.md",[],[],{"skillNo":60,"name":61,"nameZh":6,"description":62,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":63,"categories":64,"keywords":65,"publishedAt":19},"01M7DEG0BC1AY9R72J1TENYX4G","translate-text","Translate text between languages, keeping formatting and terminology.","https:\u002F\u002Fgithub.com\u002Fromeroabelleira\u002Fxeito\u002Fblob\u002Fmain\u002Fpriv\u002Fskills\u002Fbench\u002Fskills\u002Ftranslate-text\u002FSKILL.md",[],[],{"skillNo":67,"name":68,"nameZh":6,"description":69,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":70,"categories":71,"keywords":72,"publishedAt":19},"01M7DEFY61C36WVQ1Z78NEBSN0","screenshot-ui","Take screenshots of a web app's pages in a browser to check the layout.","https:\u002F\u002Fgithub.com\u002Fromeroabelleira\u002Fxeito\u002Fblob\u002Fmain\u002Fpriv\u002Fskills\u002Fbench\u002Fskills\u002Fscreenshot-ui\u002FSKILL.md",[],[],{"skillNo":74,"name":75,"nameZh":6,"description":76,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":77,"categories":78,"keywords":79,"publishedAt":19},"01M7DEFWX1JQH6BZBWG9B8NKYN","release-notes","Draft release notes from the commits since the last tag.","https:\u002F\u002Fgithub.com\u002Fromeroabelleira\u002Fxeito\u002Fblob\u002Fmain\u002Fpriv\u002Fskills\u002Fbench\u002Fskills\u002Frelease-notes\u002FSKILL.md",[],[],{"skillNo":81,"name":82,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"publishedAt":19},"01M7DEFVWS9H75PATHHS8WTF9X","pdf-extract","Extract text and tables from PDF files.","https:\u002F\u002Fgithub.com\u002Fromeroabelleira\u002Fxeito\u002Fblob\u002Fmain\u002Fpriv\u002Fskills\u002Fbench\u002Fskills\u002Fpdf-extract\u002FSKILL.md",[],[]]