[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M71JBRC85DX6JKYM70WMNXCN":3,"skill-same:01M71JBRC85DX6JKYM70WMNXCN":74},{"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":23,"publishedAt":44,"linkOnly":45,"markdownContent":46,"markdownBody":47,"frontMatter":48,"refName":51,"commitSha":52,"skillDirectory":5,"forksCount":53,"repositoryHtmlUrl":54,"repositoryHomepageUrl":55,"ownerHtmlUrl":56,"ownerBlogUrl":57,"ownerDescription":58,"updatedAt":59,"install":60,"resources":73},"01M71JBRC85DX6JKYM70WMNXCN","app-screenshots",null,"Capture and curate screenshots of a GUI desktop app for a README, landing page, app-store listing, or docs. Use when the user wants to take or refresh app screenshots — especially after a UI\u002Fredesign change. The user \"records\" the app's flows once (driving the GUI) while the agent continuously screenshots; the agent then identifies each screen, dedupes, selects, compresses, and places the right shots for the target — automatically by convention, or to the user's specific request. Triggers: \"take\u002Fupdate\u002Frefresh screenshots\", \"shoot the landing page \u002F README\", \"screenshots for the app store\", \"capture the app's screens\", \"record the app for screenshots\".","caezium\u002Fskills","caezium","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F113233555?v=4","01M71JB77ZPBAWEK009Y2W8BP7","Henry Zhang","01M71JB792NVXCWSYD27HQ0964",3,"MIT","https:\u002F\u002Fgithub.com\u002Fcaezium\u002Fskills\u002Fblob\u002Fmain\u002Fapp-screenshots\u002FSKILL.md",[18],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",[24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43],"screenshots","app screenshots","app","README","page","for","target","ui","take","curate","update","request","redesign","screen","change","listing","when","app store","desktop app","flows","2026-11-10 10:12:30",false,"---\nname: app-screenshots\ndescription: >-\n  Capture and curate screenshots of a GUI desktop app for a README, landing\n  page, app-store listing, or docs. Use when the user wants to take or refresh\n  app screenshots — especially after a UI\u002Fredesign change. The user \"records\"\n  the app's flows once (driving the GUI) while the agent continuously\n  screenshots; the agent then identifies each screen, dedupes, selects,\n  compresses, and places the right shots for the target — automatically by\n  convention, or to the user's specific request. Triggers: \"take\u002Fupdate\u002Frefresh\n  screenshots\", \"shoot the landing page \u002F README\", \"screenshots for the app\n  store\", \"capture the app's screens\", \"record the app for screenshots\".\nmetadata:\n  type: reference\n---\n\n# App screenshots — record once, agent curates\n\nTake polished, current screenshots of a GUI app and drop them where they're\nneeded (README grid, landing-page gallery, store listing, docs). The human does\nthe smallest possible amount of work: **record the flows once**. The agent does\nthe rest — capturing, identifying which screen is which, deduping, choosing the\nright frames for the target, compressing, and wiring them in.\n\nTwo modes, same recording:\n- **Auto** — agent infers what the target conventionally needs (hero, each key\n  pane, before\u002Fafter action states) and selects from the recording.\n- **Directed** — user says \"I need the Clean *running* state\" → agent finds that\n  exact frame in the recording.\n\n---\n\n## ⚠️ The one gotcha that breaks everything (read first)\n\n`screencapture -l\u003CwindowID>` and the on-screen window list (`CGWindowList…\nOnScreenOnly`) **only see a window that is composited on the *active* Space and\nnot minimized.** A window on another Desktop\u002FSpace, minimized, or fully occluded\nreturns **\"could not create image from window\"** or simply isn't listed — so a\ncapture loop silently produces nothing.\n\nThis is NOT display sleep. Symptoms look identical (blank\u002Ffailed captures) but\nthe cause is window focus\u002FSpace. Before *every* capture:\n\n1. **Foreground the target app** (`open -a`, or activate by bundle id), and\n2. confirm its window is in the on-screen list before shooting.\n\nIn a capture loop, re-activate each cycle (or verify the window id resolves and\nskip+log the cycle if it doesn't). The classic failure: launch the app detached\n(`nohup …&`) without activating, the window opens on a background Space, and the\nwhole loop captures zero frames until the user happens to bring it forward.\n\n```bash\n# Activate, then verify the window is actually on-screen before capturing.\nosascript -e 'tell application id \"com.example.App\" to activate'; sleep 1\nwinid() { swift \u002Ftmp\u002Fwinid.swift 2>\u002Fdev\u002Fnull \\\n  | awk -F'\\t' '$2==\"App\" && $4==\"layer=0\"{split($3,d,\"x\"); if(d[1]+0>600&&d[2]+0>400) print $1}' | head -1; }\nid=$(winid); [ -n \"$id\" ] && screencapture -x -l\"$id\" -o frame.png || echo \"window not on active Space — skip\"\n```\n\n---\n\n## Workflow\n\n### 1. Scope it\n- **App + how to reach each screen.** Find the launch command (or built `.app`).\n  Many apps expose a dev\u002Flaunch hook to open a specific pane non-interactively\n  (e.g. an env var like `APP_OPEN_ON_LAUNCH=\u003Cpane>`). Grep the code for one; it\n  turns static panes into scriptable, no-click captures. Sections behind a\n  segmented control or that need a button click are *not* reachable this way —\n  those go in the recording (step 2b).\n- **Target + its shot list.** READMEs want a feature grid (each pane) + action\n  states (before\u002Fafter a scan\u002Frun). Landing pages want a hero + a tab gallery.\n  Store listings want N specific sizes. Pin down the list before capturing.\n- **Check the app's current IA.** Panes get renamed, merged, or folded between\n  versions. Verify the *current* structure (don't trust an old screenshot set) —\n  if two tools folded into one hub, the old separate shots are stale.\n\n### 2. Capture\n\n**2a. Non-interactive panes (no human needed).** Launch the app once per pane\nvia the launch hook, foreground it, capture the window. Force a consistent theme\n(e.g. an appearance env var) and consistent window size so the set matches.\nSome panes auto-run work on entry (a dashboard, an auto-scan) — wait for them to\npopulate (poll the frame) before capturing.\n\n**2b. Interactive states — \"record once\".** Scans, runs, multi-step flows, and\nany state reached by clicking can't be scripted cleanly (and driving the user's\nmouse may be unavailable or unwanted). So:\n1. Launch the app and **tell the user exactly what to click** (\"open Clean → a\n   category → Scan; then Optimize → Run; enter your password if asked\").\n2. Run a **foregrounded capture loop** — a frame every ~1.5–2 s for a generous\n   window (≥2 min) — while they drive. They click; you screenshot. **Keep the\n   window foregrounded** (the gotcha above) or the loop captures nothing.\n3. The user only records **once**; you mine the frames for every state they\n   passed through.\n\n> Driving the GUI yourself (computer-use) is an option *only if available and the\n> user agrees*. Default to \"tell the user to click, you screenshot.\" Never auto-\n> trigger destructive\u002Fprivileged actions (cache deletion, maintenance needing a\n> password) for a screenshot — capture the non-destructive states (scan\u002Fpreview),\n> and let the user run + auth the rest if they want those frames.\n\n### 3. Identify + index\nFor each frame, **read it** (you can view images) and classify pane + state from\nthe on-screen text (headers, breadcrumbs, button labels) plus what you know about\nthe app. Build a frame→(feature, state) index. Cheap pre-filter: cluster frames\nby file size to find transitions (a run of identical sizes = a static screen;\nvarying sizes = an animating\u002Fin-progress screen) so you read only representatives.\n\n### 4. Select + dedupe\n- **Auto:** map the index onto the target's shot list (hero, each pane, the\n  best before\u002Fafter).\n- **Directed:** locate the user's named state in the index.\n- Pick the **clearest, most-populated** frame per state; drop blurry transitions\n  and half-loaded screens. Prefer a \"result with data\" over an empty just-started\n  frame.\n\n### 5. Process + place\n- **Compress.** Raw retina PNGs are ~2–3 MB each; `pngquant --quality 60-85\n  --strip` cuts them to ~300–500 KB with no visible loss (`brew install pngquant`\n  if missing). Match the existing assets' dimensions so the layout doesn't shift.\n- **Place + wire.** Overwrite the existing asset filenames where possible (the\n  doc keeps working), or add new ones and update the references. In a README,\n  prefer **repo-relative** image paths (`docs\u002Fassets\u002Fshot-x.png`) — GitHub\n  resolves them per-branch, so they render on the PR and after merge.\n- **Update structure if the IA changed** (drop folded tabs\u002Frows; relabel hubs).\n- Open a PR; the human reviews the composed result (you often can't render the\n  final page locally).\n\n---\n\n## Privacy — real screens show real data\nWithout a demo\u002Fsample-data mode, every shot shows the user's actual machine\n(installed apps, file\u002Ffolder names in a disk treemap, running processes, disk\nfigures). A PR is **public the moment you push.** So:\n- Confirm the user is OK publishing real data, or\n- Point data views at neutral targets (e.g. a disk analyzer at `\u002FApplications`),\n  favor metric\u002Fhero panes (no personal data), and review each frame for anything\n  too personal before it goes public.\n- A small **demo-data mode** (curated fixtures behind a flag) is the\n  professional long-term fix — clean, consistent, repeatable for every release.\n\n---\n\n## Gotchas checklist\n- [ ] Window foregrounded \u002F on the active Space before each capture (THE bug).\n- [ ] `screencapture -l\u003Cid>` for clean window-only shots (no cursor, no desktop);\n      computer-use screenshots include the whole screen + cursor.\n- [ ] Consistent theme + window size across the whole set.\n- [ ] Auto-running panes: poll until populated before shooting.\n- [ ] Interactive states: record-once loop; the user clicks, you screenshot.\n- [ ] Never auto-run destructive\u002Fprivileged actions for a shot.\n- [ ] Verify the app's *current* IA; drop stale\u002Ffolded panes.\n- [ ] Compress (pngquant) + match existing dimensions.\n- [ ] README: repo-relative image paths.\n- [ ] Real-machine data is public on push — get consent \u002F neutralize \u002F review.\n","# App screenshots — record once, agent curates\n\nTake polished, current screenshots of a GUI app and drop them where they're\nneeded (README grid, landing-page gallery, store listing, docs). The human does\nthe smallest possible amount of work: **record the flows once**. The agent does\nthe rest — capturing, identifying which screen is which, deduping, choosing the\nright frames for the target, compressing, and wiring them in.\n\nTwo modes, same recording:\n- **Auto** — agent infers what the target conventionally needs (hero, each key\n  pane, before\u002Fafter action states) and selects from the recording.\n- **Directed** — user says \"I need the Clean *running* state\" → agent finds that\n  exact frame in the recording.\n\n---\n\n## ⚠️ The one gotcha that breaks everything (read first)\n\n`screencapture -l\u003CwindowID>` and the on-screen window list (`CGWindowList…\nOnScreenOnly`) **only see a window that is composited on the *active* Space and\nnot minimized.** A window on another Desktop\u002FSpace, minimized, or fully occluded\nreturns **\"could not create image from window\"** or simply isn't listed — so a\ncapture loop silently produces nothing.\n\nThis is NOT display sleep. Symptoms look identical (blank\u002Ffailed captures) but\nthe cause is window focus\u002FSpace. Before *every* capture:\n\n1. **Foreground the target app** (`open -a`, or activate by bundle id), and\n2. confirm its window is in the on-screen list before shooting.\n\nIn a capture loop, re-activate each cycle (or verify the window id resolves and\nskip+log the cycle if it doesn't). The classic failure: launch the app detached\n(`nohup …&`) without activating, the window opens on a background Space, and the\nwhole loop captures zero frames until the user happens to bring it forward.\n\n```bash\n# Activate, then verify the window is actually on-screen before capturing.\nosascript -e 'tell application id \"com.example.App\" to activate'; sleep 1\nwinid() { swift \u002Ftmp\u002Fwinid.swift 2>\u002Fdev\u002Fnull \\\n  | awk -F'\\t' '$2==\"App\" && $4==\"layer=0\"{split($3,d,\"x\"); if(d[1]+0>600&&d[2]+0>400) print $1}' | head -1; }\nid=$(winid); [ -n \"$id\" ] && screencapture -x -l\"$id\" -o frame.png || echo \"window not on active Space — skip\"\n```\n\n---\n\n## Workflow\n\n### 1. Scope it\n- **App + how to reach each screen.** Find the launch command (or built `.app`).\n  Many apps expose a dev\u002Flaunch hook to open a specific pane non-interactively\n  (e.g. an env var like `APP_OPEN_ON_LAUNCH=\u003Cpane>`). Grep the code for one; it\n  turns static panes into scriptable, no-click captures. Sections behind a\n  segmented control or that need a button click are *not* reachable this way —\n  those go in the recording (step 2b).\n- **Target + its shot list.** READMEs want a feature grid (each pane) + action\n  states (before\u002Fafter a scan\u002Frun). Landing pages want a hero + a tab gallery.\n  Store listings want N specific sizes. Pin down the list before capturing.\n- **Check the app's current IA.** Panes get renamed, merged, or folded between\n  versions. Verify the *current* structure (don't trust an old screenshot set) —\n  if two tools folded into one hub, the old separate shots are stale.\n\n### 2. Capture\n\n**2a. Non-interactive panes (no human needed).** Launch the app once per pane\nvia the launch hook, foreground it, capture the window. Force a consistent theme\n(e.g. an appearance env var) and consistent window size so the set matches.\nSome panes auto-run work on entry (a dashboard, an auto-scan) — wait for them to\npopulate (poll the frame) before capturing.\n\n**2b. Interactive states — \"record once\".** Scans, runs, multi-step flows, and\nany state reached by clicking can't be scripted cleanly (and driving the user's\nmouse may be unavailable or unwanted). So:\n1. Launch the app and **tell the user exactly what to click** (\"open Clean → a\n   category → Scan; then Optimize → Run; enter your password if asked\").\n2. Run a **foregrounded capture loop** — a frame every ~1.5–2 s for a generous\n   window (≥2 min) — while they drive. They click; you screenshot. **Keep the\n   window foregrounded** (the gotcha above) or the loop captures nothing.\n3. The user only records **once**; you mine the frames for every state they\n   passed through.\n\n> Driving the GUI yourself (computer-use) is an option *only if available and the\n> user agrees*. Default to \"tell the user to click, you screenshot.\" Never auto-\n> trigger destructive\u002Fprivileged actions (cache deletion, maintenance needing a\n> password) for a screenshot — capture the non-destructive states (scan\u002Fpreview),\n> and let the user run + auth the rest if they want those frames.\n\n### 3. Identify + index\nFor each frame, **read it** (you can view images) and classify pane + state from\nthe on-screen text (headers, breadcrumbs, button labels) plus what you know about\nthe app. Build a frame→(feature, state) index. Cheap pre-filter: cluster frames\nby file size to find transitions (a run of identical sizes = a static screen;\nvarying sizes = an animating\u002Fin-progress screen) so you read only representatives.\n\n### 4. Select + dedupe\n- **Auto:** map the index onto the target's shot list (hero, each pane, the\n  best before\u002Fafter).\n- **Directed:** locate the user's named state in the index.\n- Pick the **clearest, most-populated** frame per state; drop blurry transitions\n  and half-loaded screens. Prefer a \"result with data\" over an empty just-started\n  frame.\n\n### 5. Process + place\n- **Compress.** Raw retina PNGs are ~2–3 MB each; `pngquant --quality 60-85\n  --strip` cuts them to ~300–500 KB with no visible loss (`brew install pngquant`\n  if missing). Match the existing assets' dimensions so the layout doesn't shift.\n- **Place + wire.** Overwrite the existing asset filenames where possible (the\n  doc keeps working), or add new ones and update the references. In a README,\n  prefer **repo-relative** image paths (`docs\u002Fassets\u002Fshot-x.png`) — GitHub\n  resolves them per-branch, so they render on the PR and after merge.\n- **Update structure if the IA changed** (drop folded tabs\u002Frows; relabel hubs).\n- Open a PR; the human reviews the composed result (you often can't render the\n  final page locally).\n\n---\n\n## Privacy — real screens show real data\nWithout a demo\u002Fsample-data mode, every shot shows the user's actual machine\n(installed apps, file\u002Ffolder names in a disk treemap, running processes, disk\nfigures). A PR is **public the moment you push.** So:\n- Confirm the user is OK publishing real data, or\n- Point data views at neutral targets (e.g. a disk analyzer at `\u002FApplications`),\n  favor metric\u002Fhero panes (no personal data), and review each frame for anything\n  too personal before it goes public.\n- A small **demo-data mode** (curated fixtures behind a flag) is the\n  professional long-term fix — clean, consistent, repeatable for every release.\n\n---\n\n## Gotchas checklist\n- [ ] Window foregrounded \u002F on the active Space before each capture (THE bug).\n- [ ] `screencapture -l\u003Cid>` for clean window-only shots (no cursor, no desktop);\n      computer-use screenshots include the whole screen + cursor.\n- [ ] Consistent theme + window size across the whole set.\n- [ ] Auto-running panes: poll until populated before shooting.\n- [ ] Interactive states: record-once loop; the user clicks, you screenshot.\n- [ ] Never auto-run destructive\u002Fprivileged actions for a shot.\n- [ ] Verify the app's *current* IA; drop stale\u002Ffolded panes.\n- [ ] Compress (pngquant) + match existing dimensions.\n- [ ] README: repo-relative image paths.\n- [ ] Real-machine data is public on push — get consent \u002F neutralize \u002F review.\n",{"name":5,"metadata":49,"description":7},{"type":50},"reference","main","6ef866230c1ba3459fd08fc65141dfd37fffd41a",0,"https:\u002F\u002Fgithub.com\u002Fcaezium\u002Fskills","https:\u002F\u002Fskills.henryzh.dev","https:\u002F\u002Fgithub.com\u002Fcaezium","henryzh.dev","quant, data science, full-stack, robotics.\r\nbuilding cool stuff","2026-08-24 19:21:44",{"source":8,"skillName":5,"commands":61,"prompt":71,"zipAvailable":72},[62,65,68],{"tool":63,"command":64},"npx","npx skills add caezium\u002Fskills --skill app-screenshots",{"tool":66,"command":67},"bunx","bunx skills add caezium\u002Fskills --skill app-screenshots",{"tool":69,"command":70},"pnpm","pnpm dlx skills add caezium\u002Fskills --skill app-screenshots","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fcaezium\u002Fskills\u002Fblob\u002Fmain\u002Fapp-screenshots\u002FSKILL.md",true,[],[75,95,123,141,162,184],{"skillNo":76,"name":77,"nameZh":6,"description":78,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":79,"categories":80,"keywords":89,"publishedAt":44},"01M71JT4RYCAS18RF6003YYR1D","writing-for-agents","Writing documents for agents. Use when creating or editing skills, or modifying AGENTS.md or CLAUDE.md.","https:\u002F\u002Fgithub.com\u002Fcaezium\u002Fskills\u002Fblob\u002Fmain\u002Fwriting-for-agents\u002FSKILL.md",[81,84],{"categoryNo":82,"categoryName":83,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000018","写作与文案",{"categoryNo":85,"categoryName":86,"taxonomyType":87,"taxonomyTypeDesc":88},"01JBSEEDCAT000000000000036","内容创作者","PROFESSION","适用职业",[90,29,91,92,93,94,40],"writing for agents","writing","editing","use","documents",{"skillNo":96,"name":97,"nameZh":6,"description":98,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":99,"categories":100,"keywords":104,"publishedAt":44},"01M71JT0GES4AZ5ADH8EADNN1Z","wrangler","Cloudflare Workers CLI for deploying, developing, and managing Workers, KV, R2, D1, Vectorize, Hyperdrive, Workers AI, Containers, Queues, Workflows, Pipelines, and Secrets Store. Load before running wrangler commands to ensure correct syntax and best practices. Biases towards retrieval from Cloudflare docs over pre-trained knowledge.","https:\u002F\u002Fgithub.com\u002Fcaezium\u002Fskills\u002Fblob\u002Fmain\u002Fwrangler\u002FSKILL.md",[101],{"categoryNo":102,"categoryName":103,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000010","后端与接口",[97,105,106,107,108,109,110,111,112,113,114,115,116,29,117,118,119,120,121,122],"cloudflare","d1","knowledge","workflows","cli","secrets","commands","syntax","cloudflare workers","r2","queues","running","containers","correct","load","workers","best practices","store",{"skillNo":124,"name":125,"nameZh":6,"description":126,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":127,"categories":128,"keywords":129,"publishedAt":44},"01M71JSWX4R203M32K9A18AAW8","workers-best-practices","Reviews and authors Cloudflare Workers code against production best practices. Load when writing new Workers, reviewing Worker code, configuring wrangler.jsonc, or checking for common Workers anti-patterns (streaming, floating promises, global state, secrets, bindings, observability). Biases towards retrieval from Cloudflare docs over pre-trained knowledge.","https:\u002F\u002Fgithub.com\u002Fcaezium\u002Fskills\u002Fblob\u002Fmain\u002Fworkers-best-practices\u002FSKILL.md",[],[130,120,121,131,132,133,29,105,134,107,135,97,91,119,40,136,137,138,139,140],"workers best practices","observability","reviews","promises","worker","new","code","common","streaming","state","anti-patterns",{"skillNo":142,"name":143,"nameZh":6,"description":144,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":145,"categories":146,"keywords":147,"publishedAt":44},"01M71JST6P88MXGT2MWDTPJGEG","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\u002Fcaezium\u002Fskills\u002Fblob\u002Fmain\u002Fwizard\u002FSKILL.md",[],[143,116,148,29,149,150,151,152,93,153,40,154,155,156,157,158,159,160,161,110],"setting","up","ci","provisioning infrastructure","only","invoke","Bash","one","interactive","can","credentials","dashboard","generate","human",{"skillNo":163,"name":164,"nameZh":6,"description":165,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":166,"sourceHtmlUrl":167,"categories":168,"keywords":169,"publishedAt":44},"01M71JSQ95S7BQ61DAPN95X2W4","why","Use for 'why does X work this way', 'why we picked Y', design rationale, regressions, postmortems, or data-backed thresholds. Discovers available evidence sources, searches each relevant category, and returns a cited read on decisions and tradeoffs. Use how for runtime behavior.","https:\u002F\u002Fgithub.com\u002Fcaezium\u002Fskills\u002Fblob\u002Fmain\u002Fwhy\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fcaezium\u002Fskills\u002Fblob\u002Fmain\u002Fwhy\u002FSKILL.md",[],[164,170,171,172,173,29,174,175,93,176,177,178,179,180,181,182,183],"sources","thresholds","evidence","relevant","runtime","design","decisions","work","how","behavior","read","data","rationale","returns",{"skillNo":185,"name":186,"nameZh":6,"description":187,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":188,"categories":189,"keywords":191,"publishedAt":44},"01M71JSMHFVSH8HK8GXN6YAX6J","week-prep","Generate a briefing with calendar, tasks, and attention flags. Use for \"week prep\", \"briefing\", or \"what's on this week\".","https:\u002F\u002Fgithub.com\u002Fcaezium\u002Fskills\u002Fblob\u002Fmain\u002Fweek-prep\u002FSKILL.md",[190],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[192,193,194,195,196,29,197,198,199,200,93,154,201,202,203,160,204,180,205,206],"week prep","prep","deep","attention","get","search","messages","briefing","MCP","tasks","with","slack","calendar","flags","grep"]