[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7CW4A9NMAS3TC6ZVJ0R4Q13":3,"skill-same:01M7CW4A9NMAS3TC6ZVJ0R4Q13":45},{"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":29,"updatedAt":30,"install":31,"resources":44},"01M7CW4A9NMAS3TC6ZVJ0R4Q13","skill-router",null,"Search a local corpus of ~6,000 curated agent skills (testing, frontend, backend, cloud, data, writing, security, research, media) BEFORE doing any concrete task. Use at the START of any build\u002Ffix\u002Ftest\u002Fdeploy\u002Fconfigure\u002Frefactor\u002Freview request, unprompted - call skills_search with the task's goal, read the best-matching playbook with skills_read, and follow it. Default is ROUTE; skip only for pure chat.","ryasrk\u002Fdsh-expert-agents","ryasrk","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F115061102?v=4","01M7CW079PK1VJJZPP5KXWFRXN","Ryas Rafi Karim","01M7CW07AKNEJV8913PZB163JK",0,"MIT","https:\u002F\u002Fgithub.com\u002Fryasrk\u002Fdsh-expert-agents\u002Fblob\u002Fmain\u002Fassets\u002Fskills\u002Fskill-router\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: skill-router\ndescription: Search a local corpus of ~6,000 curated agent skills (testing, frontend, backend, cloud, data, writing, security, research, media) BEFORE doing any concrete task. Use at the START of any build\u002Ffix\u002Ftest\u002Fdeploy\u002Fconfigure\u002Frefactor\u002Freview request, unprompted - call skills_search with the task's goal, read the best-matching playbook with skills_read, and follow it. Default is ROUTE; skip only for pure chat.\n---\n\n# Skill router\n\nYou have access to a local corpus of **~6,000 specialized agent skills** that\nis NOT in your catalog. It is vector-indexed and searchable in about a second.\nA vetted playbook almost always beats improvising from general knowledge - so\nsearch BEFORE planning, not after.\n\n## When to route\n\nRoute by reading the conversation. A lookup is triggered by the SHAPE of the\nmessage, not by the user naming a skill:\n\n- The user describes something they want built, fixed, tested, deployed,\n  configured, refactored, reviewed, or explained with intent to act.\n- The user names a tool, framework, library, cloud, or database.\n- The user asks \"how do I \u002F can we \u002F please do\".\n- The conversation is mid-task and the next step is a concrete action.\n\nSkip only when the turn is purely conversational: a greeting, thanks, a\nclarification question, a status report - or you are already following a\nskill loaded this turn.\n\nDefault is ROUTE. If unsure whether a turn counts, route anyway - a miss\ncosts one second. Silence after an actionable message is the failure mode\nthis skill exists to prevent.\n\n## How to search\n\nWrite the query the way you would explain the task to a colleague - a goal,\nnot a keyword list. Semantic matching does the rest.\n\nCall the `skills_search` tool with `query` (and `k` if you want more than\n5 hits). Each result carries `path`, `score` (0-1), `name`, and a one-line\n`description`.\n\nIf this host has no `skills_search` tool, run the bundled search CLI once:\nfeed it the JSON query `{\"query\":\"\u003Cgoal in plain words>\",\"k\":5}` on its\nstandard input (it reads stdin and writes JSON to stdout; adapt the stdin\u002Fpipe\ninvocation to whatever shell this host provides — on Linux\u002FmacOS that is bash,\non Windows it is PowerShell):\n```\nnode \"\u002Fhome\u002Fryasr\u002F.dsh\u002Fprofiles\u002Fweb\u002Fnode_modules\u002Fdsh-awesome-skills\u002Flib\u002Fquery.js\"\n```\nThe CLI answers with a JSON object whose `results` are the ranked hits.\n\n### Reading the scores\n\nCalibrated on labeled task queries against this corpus: strong-band top hits\nwere relevant nearly every time; mid-band hits are plausible but need a skim;\nweak-band hits rarely help. Trust the bands:\n\n- **0.7+** strong match - read it.\n- **0.4-0.7** plausible - read the top 2-3 descriptions, not\n  just the first line; the best playbook for the task is often at rank 2-3.\n- **\u003C 0.4** weak - likely no good playbook for this exact task.\n\n### When one query is not enough\n\n- **Broad request** (\"build me a dashboard\") → run 2-3 queries for its\n  distinct facets (e.g. \"frontend dashboard layout\", \"charts data\n  visualization\", \"deploy static site\") and take the best hit from each.\n- **First results all score \u003C 0.4** → rephrase ONCE with different\n  wording (\"auth\" → \"login session cookie\"). A second miss means the corpus\n  likely has nothing; move on without guilt.\n- **Multiple hits with the same trailing name** from different sources\n  (e.g. several `tdd` skills, or duplicates across\n  `a5c-ai\u002Fbabysitter\u002F\u003Cdomain>\u002F` specializations) → they are alternative\n  takes. The full `path` distinguishes them. Prefer the one whose\n  description fits the user's stack; when they look equivalent, prefer the\n  higher score and read just one.\n\n### Decision points, don't drift\n\n- Task is one clear domain and a hit scores 0.7+ → read exactly one\n  skill and follow it.\n- Task genuinely spans domains (e.g. \"add OAuth to my FastAPI app\") → read at\n  most two skills (one per domain), name which is primary, and follow the\n  primary for the overall shape.\n\n## How to read a hit\n\nCall the `skills_read` tool with the hit's `path` (**never** the display\n`name` - paths are unique) to load its SKILL.md. If this host has no\n`skills_read` tool, read the file directly with the host's file-read\ncapability at `\u002Fhome\u002Fryasr\u002F.dsh\u002Fawesome-skills\u002Fskills\u002F\u003Cpath>\u002FSKILL.md`.\n\nA skill's directory holds its complete playbook: reference files, templates,\nexamples, and scripts live beside the SKILL.md. When the SKILL.md says \"read\ntests.md\" or points at `references\u002F` or `scripts\u002F`, load those with\n`skills_read` (path + `file`) too - they are part of the skill, not optional\ndecoration. If `skills_read` is unavailable, list the directory\n`\u002Fhome\u002Fryasr\u002F.dsh\u002Fawesome-skills\u002Fskills\u002F\u003Cpath>\u002F` with the host's file browser to see its contents.\n\n## Rules\n\n- Read before acting: search → read the match → then plan. Quoting a skill's\n  title from memory is not following it.\n- Never paste a whole skill file into the chat - read it, then act on it.\n- Follow the playbook, adapt the details. If a step genuinely does not fit\n  the user's context, say so and adapt rather than silently skipping.\n- If nothing relevant comes back after one rephrase, continue normally. A\n  miss costs one second and is not a failure.\n","# Skill router\n\nYou have access to a local corpus of **~6,000 specialized agent skills** that\nis NOT in your catalog. It is vector-indexed and searchable in about a second.\nA vetted playbook almost always beats improvising from general knowledge - so\nsearch BEFORE planning, not after.\n\n## When to route\n\nRoute by reading the conversation. A lookup is triggered by the SHAPE of the\nmessage, not by the user naming a skill:\n\n- The user describes something they want built, fixed, tested, deployed,\n  configured, refactored, reviewed, or explained with intent to act.\n- The user names a tool, framework, library, cloud, or database.\n- The user asks \"how do I \u002F can we \u002F please do\".\n- The conversation is mid-task and the next step is a concrete action.\n\nSkip only when the turn is purely conversational: a greeting, thanks, a\nclarification question, a status report - or you are already following a\nskill loaded this turn.\n\nDefault is ROUTE. If unsure whether a turn counts, route anyway - a miss\ncosts one second. Silence after an actionable message is the failure mode\nthis skill exists to prevent.\n\n## How to search\n\nWrite the query the way you would explain the task to a colleague - a goal,\nnot a keyword list. Semantic matching does the rest.\n\nCall the `skills_search` tool with `query` (and `k` if you want more than\n5 hits). Each result carries `path`, `score` (0-1), `name`, and a one-line\n`description`.\n\nIf this host has no `skills_search` tool, run the bundled search CLI once:\nfeed it the JSON query `{\"query\":\"\u003Cgoal in plain words>\",\"k\":5}` on its\nstandard input (it reads stdin and writes JSON to stdout; adapt the stdin\u002Fpipe\ninvocation to whatever shell this host provides — on Linux\u002FmacOS that is bash,\non Windows it is PowerShell):\n```\nnode \"\u002Fhome\u002Fryasr\u002F.dsh\u002Fprofiles\u002Fweb\u002Fnode_modules\u002Fdsh-awesome-skills\u002Flib\u002Fquery.js\"\n```\nThe CLI answers with a JSON object whose `results` are the ranked hits.\n\n### Reading the scores\n\nCalibrated on labeled task queries against this corpus: strong-band top hits\nwere relevant nearly every time; mid-band hits are plausible but need a skim;\nweak-band hits rarely help. Trust the bands:\n\n- **0.7+** strong match - read it.\n- **0.4-0.7** plausible - read the top 2-3 descriptions, not\n  just the first line; the best playbook for the task is often at rank 2-3.\n- **\u003C 0.4** weak - likely no good playbook for this exact task.\n\n### When one query is not enough\n\n- **Broad request** (\"build me a dashboard\") → run 2-3 queries for its\n  distinct facets (e.g. \"frontend dashboard layout\", \"charts data\n  visualization\", \"deploy static site\") and take the best hit from each.\n- **First results all score \u003C 0.4** → rephrase ONCE with different\n  wording (\"auth\" → \"login session cookie\"). A second miss means the corpus\n  likely has nothing; move on without guilt.\n- **Multiple hits with the same trailing name** from different sources\n  (e.g. several `tdd` skills, or duplicates across\n  `a5c-ai\u002Fbabysitter\u002F\u003Cdomain>\u002F` specializations) → they are alternative\n  takes. The full `path` distinguishes them. Prefer the one whose\n  description fits the user's stack; when they look equivalent, prefer the\n  higher score and read just one.\n\n### Decision points, don't drift\n\n- Task is one clear domain and a hit scores 0.7+ → read exactly one\n  skill and follow it.\n- Task genuinely spans domains (e.g. \"add OAuth to my FastAPI app\") → read at\n  most two skills (one per domain), name which is primary, and follow the\n  primary for the overall shape.\n\n## How to read a hit\n\nCall the `skills_read` tool with the hit's `path` (**never** the display\n`name` - paths are unique) to load its SKILL.md. If this host has no\n`skills_read` tool, read the file directly with the host's file-read\ncapability at `\u002Fhome\u002Fryasr\u002F.dsh\u002Fawesome-skills\u002Fskills\u002F\u003Cpath>\u002FSKILL.md`.\n\nA skill's directory holds its complete playbook: reference files, templates,\nexamples, and scripts live beside the SKILL.md. When the SKILL.md says \"read\ntests.md\" or points at `references\u002F` or `scripts\u002F`, load those with\n`skills_read` (path + `file`) too - they are part of the skill, not optional\ndecoration. If `skills_read` is unavailable, list the directory\n`\u002Fhome\u002Fryasr\u002F.dsh\u002Fawesome-skills\u002Fskills\u002F\u003Cpath>\u002F` with the host's file browser to see its contents.\n\n## Rules\n\n- Read before acting: search → read the match → then plan. Quoting a skill's\n  title from memory is not following it.\n- Never paste a whole skill file into the chat - read it, then act on it.\n- Follow the playbook, adapt the details. If a step genuinely does not fit\n  the user's context, say so and adapt rather than silently skipping.\n- If nothing relevant comes back after one rephrase, continue normally. A\n  miss costs one second and is not a failure.\n",{"name":5,"description":7},"main","b608340412f02546f13b1b29976daf33bec72bef","assets\u002Fskills\u002Fskill-router","https:\u002F\u002Fgithub.com\u002Fryasrk\u002Fdsh-expert-agents","https:\u002F\u002Fgithub.com\u002Fryasrk","Passionate about Machine Learning, Deep Learning, and LLMs. Skilled in Computer Vision, NLP, and AI model development.","2026-09-29 12:47:36",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add ryasrk\u002Fdsh-expert-agents --skill skill-router",{"tool":37,"command":38},"bunx","bunx skills add ryasrk\u002Fdsh-expert-agents --skill skill-router",{"tool":40,"command":41},"pnpm","pnpm dlx skills add ryasrk\u002Fdsh-expert-agents --skill skill-router","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fryasrk\u002Fdsh-expert-agents\u002Fblob\u002Fmain\u002Fassets\u002Fskills\u002Fskill-router\u002FSKILL.md",true,[],[46,53,60,67,74,81],{"skillNo":47,"name":48,"nameZh":6,"description":49,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":50,"categories":51,"keywords":52,"publishedAt":19},"01M7CW5A9FJK1C6BQKKMYHKWBD","writing-shape","Writing, exploit: shape raw material into an article, paragraph by paragraph.","https:\u002F\u002Fgithub.com\u002Fryasrk\u002Fdsh-expert-agents\u002Fblob\u002Fmain\u002Fassets\u002Fskills\u002Fwriting-shape\u002FSKILL.md",[],[],{"skillNo":54,"name":55,"nameZh":6,"description":56,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":57,"categories":58,"keywords":59,"publishedAt":19},"01M7CW58T8BYGMCTKQW60DFM7K","writing-fragments","Writing, explore: mine raw fragments, no structure yet.","https:\u002F\u002Fgithub.com\u002Fryasrk\u002Fdsh-expert-agents\u002Fblob\u002Fmain\u002Fassets\u002Fskills\u002Fwriting-fragments\u002FSKILL.md",[],[],{"skillNo":61,"name":62,"nameZh":6,"description":63,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":64,"categories":65,"keywords":66,"publishedAt":19},"01M7CW56PXKJ31Y8AGJ31W1SHT","writing-for-agents","Writing documents for agents. Use when creating or editing skills, or modifying AGENTS.md or CLAUDE.md.","https:\u002F\u002Fgithub.com\u002Fryasrk\u002Fdsh-expert-agents\u002Fblob\u002Fmain\u002Fassets\u002Fskills\u002Fwriting-for-agents\u002FSKILL.md",[],[],{"skillNo":68,"name":69,"nameZh":6,"description":70,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":71,"categories":72,"keywords":73,"publishedAt":19},"01M7CW54P3DQJKVA9GBQBMXE00","writing-beats","Writing, exploit; assemble raw material into a journey of beats, grounding each term before a beat leans on it.","https:\u002F\u002Fgithub.com\u002Fryasrk\u002Fdsh-expert-agents\u002Fblob\u002Fmain\u002Fassets\u002Fskills\u002Fwriting-beats\u002FSKILL.md",[],[],{"skillNo":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":19},"01M7CW50RQRS31BHTKJ7JZM21A","wizard","Generate an interactive bash wizard that walks a human through steps only they can perform. Use when provisioning infrastructure, setting up credentials or CI secrets, walking an unfamiliar third-party dashboard, or running a one-off migration or cutover. Don't invoke this for steps the agent can perform itself.","https:\u002F\u002Fgithub.com\u002Fryasrk\u002Fdsh-expert-agents\u002Fblob\u002Fmain\u002Fassets\u002Fskills\u002Fwizard\u002FSKILL.md",[],[],{"skillNo":82,"name":83,"nameZh":6,"description":84,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":85,"categories":86,"keywords":87,"publishedAt":19},"01M7CW4Y3E85C06AZD2VANK40Q","wayfinder","Plan a huge chunk of work (more than one agent session can hold) as a shared map of decision tickets on your issue tracker, and resolve them one at a time until the way to the destination is clear.","https:\u002F\u002Fgithub.com\u002Fryasrk\u002Fdsh-expert-agents\u002Fblob\u002Fmain\u002Fassets\u002Fskills\u002Fwayfinder\u002FSKILL.md",[],[]]