Prompt library changedprompt-library
Nearest release: v2.1.287, published 10 hours before upstream edited the page. Shown because the two are within 24 hours of each other. Nothing here says the release caused the edit.
Upstream edited this page at 2 Oct 2026 03:23 UTC, give or take a minute or two: the time comes from Anthropic’s own sitemap rather than from a commit. This site recorded the change at 2 Oct 2026 03:37 UTC.
Upstream edited
Recorded here
Lines+277added
Lines−54removed
From line
618
where the diff opens
First seen
14 Aug 2026
this site's first read of the page
Recorded edits7to this page, all time
The whole hunk
from line 618, old and new numbered
/
from line 618
618618 const m = p.slice(base.length).match(/^\/([a-z]{2}(?:-[A-Z]{2})?)\//);
619619 const locale = m ? m[1] : 'en';
620620 return href => {
621 if (!href || href[0] !== '/' || href[1] === '/') return href;
621 if (!href) return undefined;
622 if (href[0] === '#' || href.startsWith('https://')) return href;
623 if (!(/^\/[A-Za-z0-9]/).test(href)) return undefined;
622624 return base + (href.startsWith('/en/') ? '/' + locale + href.slice(3) : href);
623625 };
624626 }, []);
from line 669
667669 const assemble = p => p.prompt.replace(/\{(\w+)\}/g, (_, k) => fillOf(p, k) || p.slots && p.slots[k] || k);
668670 const preview = p => p.prompt.replace(/\{(\w+)\}/g, (_, k) => p.slots && p.slots[k] || k);
669671 const bodyText = p => preview(p) + ' ' + p.teaches.replace(/\[([^\]]+)\]\([^)]+\)/g, '$1') + ' ' + (p.next || '');
670 const widthFor = s => (s || '').length + 3 + 'ch';
672 const WIDE_RE = /[\u1100-\u115F\u2E80-\uA4CF\uAC00-\uD7A3\uF900-\uFAFF\uFE30-\uFE4F\uFF00-\uFF60\uFFE0-\uFFE6]/g;
673 const widthFor = s => {
674 const t = typeof s === 'string' ? s : '';
675 return t.length + (t.match(WIDE_RE) || []).length + 3 + 'ch';
676 };
671677 const ql = q.trim().toLowerCase();
672678 const toggleTag = k => {
673679 setStart(false);
from line 1100
10941100 "get-oriented-in-a": {
10951101 title: "Get oriented in a new repository",
10961102 teaches: "Describe what you want to know, not which files to read. Claude explores the project on its own and returns a summary of how it fits together.",
1097 next: "Run `/init` to set up `CLAUDE.md` so Claude remembers this every session"
1103 next: "Run `/init` to set up `CLAUDE.md` so Claude remembers this every session",
1104 prompt: "give me an overview of this codebase: architecture, key directories, and how the pieces connect"
10981105 },
10991106 "explain-unfamiliar-code": {
11001107 title: "Explain unfamiliar code",
11011108 teaches: "Name the file and say what format you want the answer in. Swap the HTML page for a diagram, bullet points, or whatever fits how you learn.",
1102 next: "Set an output style so Claude always explains in your preferred format"
1109 next: "Set an output style so Claude always explains in your preferred format",
1110 prompt: "explain what {path} does and how data flows through it. write it up as {format}",
1111 slots: {
1112 path: "src/scheduler/queue.ts",
1113 format: "an HTML page with a diagram, then open it in my browser"
1114 }
11031115 },
11041116 "find-where-something-happens": {
11051117 title: "Find where something happens",
1106 teaches: "Search by behavior instead of by filename. The search works even when you don't know what the file is called or which directory it lives in."
1118 teaches: "Search by behavior instead of by filename. The search works even when you don't know what the file is called or which directory it lives in.",
1119 prompt: "where do we {behavior}?",
1120 slots: {
1121 behavior: "validate uploaded file types"
1122 }
11071123 },
11081124 "see-what-depends-on": {
11091125 title: "Check what breaks before you delete",
1110 teaches: "Ask before you remove anything. The list of callers and downstream effects tells you whether you're looking at a one-line cleanup or a change you need to coordinate."
1126 teaches: "Ask before you remove anything. The list of callers and downstream effects tells you whether you're looking at a one-line cleanup or a change you need to coordinate.",
1127 prompt: "what would break if I deleted {target}?",
1128 slots: {
1129 target: "the retryWithBackoff helper"
1130 }
11111131 },
11121132 "trace-how-code-evolved": {
11131133 title: "Trace how code evolved",
1114 teaches: "Point at commit history when the question is why, not what. Claude reads the log and blame for whatever version control you use and explains the decisions behind the current implementation."
1134 teaches: "Point at commit history when the question is why, not what. Claude reads the log and blame for whatever version control you use and explains the decisions behind the current implementation.",
1135 prompt: "look through the commit history of {path} and summarize how it evolved and why",
1136 slots: {
1137 path: "internal/auth/session.go"
1138 }
11151139 },
11161140 "scope-a-change-before": {
11171141 title: "Scope a change before you start",
1118 teaches: "Size the work before you commit it to a roadmap. The file list tells you whether you're looking at one component or a cross-cutting change."
1142 teaches: "Size the work before you commit it to a roadmap. The file list tells you whether you're looking at one component or a cross-cutting change.",
1143 prompt: "which files would I need to touch to {change}?",
1144 slots: {
1145 change: "add a dark mode toggle to settings"
1146 }
11191147 },
11201148 "ask-the-codebase-a": {
11211149 title: "Ask the codebase a product question",
11221150 teaches: "State your role so the answer is pitched at the right level. Claude explains what the product actually does from the source code, without you needing to read it.",
1123 next: "Set an output style so Claude always pitches answers at this level"
1151 next: "Set an output style so Claude always pitches answers at this level",
1152 prompt: "I am a {role}. walk me through what happens when a user {action}, from the UI down to the result",
1153 slots: {
1154 role: "PM",
1155 action: "clicks Export to PDF"
1156 }
11241157 },
11251158 "plan-a-multi-file": {
11261159 title: "Plan a multi-file change before touching code",
1127 teaches: "Adding \"don't edit yet\" separates exploration from changes, so you see the approach before any code moves. To make plan-first the default on every prompt, press Shift+Tab for [plan mode](/docs/en/permission-modes#analyze-before-you-edit-with-plan-mode)."
1160 teaches: "Adding \"don't edit yet\" separates exploration from changes, so you see the approach before any code moves. To make plan-first the default on every prompt, press Shift+Tab for [plan mode](/docs/en/permission-modes#analyze-before-you-edit-with-plan-mode).",
1161 prompt: "plan how to refactor the {target} to {goal}. list the files you would change, but don't edit anything yet",
1162 slots: {
1163 target: "payment module",
1164 goal: "support multiple currencies"
1165 }
11281166 },
11291167 "draft-a-spec-by": {
11301168 title: "Draft a spec by interview",
11311169 teaches: "Ask to be interviewed instead of writing the spec yourself. Claude asks you structured questions until the requirements are complete, then writes the result to a file.",
1132 next: "Save your interview questions as a `/spec` skill so every spec starts the same way"
1170 next: "Save your interview questions as a `/spec` skill so every spec starts the same way",
1171 prompt: "I want to build {feature}. interview me about implementation, UX, edge cases, and tradeoffs until we have covered everything, then write the spec to SPEC.md",
1172 slots: {
1173 feature: "per-workspace rate limits"
1174 }
11331175 },
11341176 "turn-a-meeting-into": {
11351177 title: "Turn a meeting into tickets",
11361178 teaches: "Skip the transcription step. Claude pulls action items from the unstructured input and writes them straight into your tracker via [MCP](/docs/en/mcp), so you review the tickets, not the transcript.",
1137 next: "Save this as a `/tickets` skill"
1179 next: "Save this as a `/tickets` skill",
1180 prompt: "read {input} and write up the action items, then create a {tracker} ticket for each with acceptance criteria",
1181 slots: {
1182 input: "@meeting-notes.md",
1183 tracker: "Linear"
1184 }
11381185 },
11391186 "map-edge-cases-before": {
11401187 title: "Map edge cases before building",
1141 teaches: "Ask for what's missing, not what's there. Claude lists the error states, empty states, and edge cases a happy-path design tends to skip."
1188 teaches: "Ask for what's missing, not what's there. Claude lists the error states, empty states, and edge cases a happy-path design tends to skip.",
1189 prompt: "list the error states, empty states, and edge cases for {feature} that the design needs to cover",
1190 slots: {
1191 feature: "the file upload flow"
1192 }
11421193 },
11431194 "turn-a-mockup-into": {
11441195 title: "Turn a mockup into a working prototype",
1145 teaches: "A clickable prototype answers questions a static mockup can't. Hand the working code to engineering instead of explaining the interactions in a doc."
1196 teaches: "A clickable prototype answers questions a static mockup can't. Hand the working code to engineering instead of explaining the interactions in a doc.",
1197 prompt: "here is a mockup. build a working prototype I can click through, matching the layout and states shown"
11461198 },
11471199 "implement-from-a-screenshot": {
11481200 title: "Implement from a screenshot and self-check",
11491201 teaches: "This gives Claude a verification loop: it renders, compares against the source image, and iterates without you pointing out each gap.",
1150 next: "Use `/goal` to keep Claude iterating toward matching screenshots"
1202 next: "Use `/goal` to keep Claude iterating toward matching screenshots",
1203 prompt: "implement this design, then take a screenshot of the result, compare it to the original, and fix any differences"
11511204 },
11521205 "follow-an-existing-pattern": {
11531206 title: "Follow an existing pattern",
11541207 teaches: "Point at code you already like. Without a reference, Claude defaults to general best practices. With one, it matches the conventions your codebase actually uses.",
1155 next: "Ask Claude to write the pattern it followed into `CLAUDE.md` so future sessions match it without the reference"
1208 next: "Ask Claude to write the pattern it followed into `CLAUDE.md` so future sessions match it without the reference",
1209 prompt: "look at how {example} is implemented to understand the pattern, then build {new} the same way",
1210 slots: {
1211 example: "the GitHub webhook handler",
1212 new: "a Stripe webhook handler"
1213 }
11561214 },
11571215 "add-a-small-well": {
11581216 title: "Add a small, well-defined feature",
1159 teaches: "State the inputs and outputs, not how to build it. Claude finds where similar code lives and adds yours alongside it."
1217 teaches: "State the inputs and outputs, not how to build it. Claude finds where similar code lives and adds yours alongside it.",
1218 prompt: "add a {endpoint} endpoint that returns {payload}",
1219 slots: {
1220 endpoint: "/health",
1221 payload: "the app version and uptime"
1222 }
11601223 },
11611224 "build-a-small-internal": {
11621225 title: "Build a small internal tool from scratch",
1163 teaches: "You don't need a project, a framework, or a build step. Describe the tool and ask Claude to open it so you see it working immediately."
1226 teaches: "You don't need a project, a framework, or a build step. Describe the tool and ask Claude to open it so you see it working immediately.",
1227 prompt: "create a {tool} using HTML, CSS, and vanilla JavaScript, then open it in my browser",
1228 slots: {
1229 tool: "drag-and-drop Kanban board with three columns"
1230 }
11641231 },
11651232 "work-an-issue-end": {
11661233 title: "Work an issue end to end",
1167 teaches: "Give the issue number, not a summary. Claude reads the full ticket itself, so requirements you'd forget to mention come through, and it validates the change before reporting back."
1234 teaches: "Give the issue number, not a summary. Claude reads the full ticket itself, so requirements you'd forget to mention come through, and it validates the change before reporting back.",
1235 prompt: "read issue #{issue}, implement the fix, and run the tests",
1236 slots: {
1237 issue: "312"
1238 }
11681239 },
11691240 "find-and-update-copy": {
11701241 title: "Find and update copy across the codebase",
1171 teaches: "Ask for variants and say what to skip. Claude finds phrasings a literal search would miss and leaves test fixtures and history untouched, so you review only the copy users actually see."
1242 teaches: "Ask for variants and say what to skip. Claude finds phrasings a literal search would miss and leaves test fixtures and history untouched, so you review only the copy users actually see.",
1243 prompt: "find every place we say \"{copy}\" or a close variant, show me each one in context, then update them all to \"{new}\". leave tests and the changelog alone",
1244 slots: {
1245 copy: "Sign up free",
1246 new: "Start free trial"
1247 }
11721248 },
11731249 "draft-from-past-examples": {
11741250 title: "Draft a document from past examples",
11751251 teaches: "Point at a folder of finished work instead of describing your style. Claude learns the structure and voice from what you've already shipped, so the first draft reads like one of yours.",
1176 next: "Save the voice as a skill so every draft starts there"
1252 next: "Save the voice as a skill so every draft starts there",
1253 prompt: "read the {examples} in {folder} to learn the structure and voice, then draft a new one for {topic}",
1254 slots: {
1255 examples: "privacy impact assessments",
1256 folder: "legal/pia/",
1257 topic: "the new analytics integration"
1258 }
11771259 },
11781260 "write-tests-run-them": {
11791261 title: "Write tests, run them, fix failures",
11801262 teaches: "Ask for write, run, and fix together so Claude iterates without stopping for instructions.",
1181 next: "Run `/init` so Claude learns your test command automatically"
1263 next: "Run `/init` so Claude learns your test command automatically",
1264 prompt: "write tests for {path}, run them, and fix any failures",
1265 slots: {
1266 path: "app/parsers/feed.py"
1267 }
11821268 },
11831269 "drive-implementation-from-tests": {
11841270 title: "Drive implementation from tests",
1185 teaches: "Test-driven development: the tests define when the work is complete, and Claude iterates on the implementation until they pass."
1271 teaches: "Test-driven development: the tests define when the work is complete, and Claude iterates on the implementation until they pass.",
1272 prompt: "write tests for {feature} first, then implement it until they pass",
1273 slots: {
1274 feature: "the password reset flow"
1275 }
11861276 },
11871277 "fill-gaps-from-a": {
11881278 title: "Fill gaps from a coverage report",
11891279 teaches: "Point at the coverage report instead of guessing what's untested. Claude reads the actual numbers and writes tests for the files that need them most.",
1190 next: "Set this as a `/goal` so Claude keeps writing tests toward the coverage target"
1280 next: "Set this as a `/goal` so Claude keeps writing tests toward the coverage target",
1281 prompt: "read {report} and add tests for the lowest-covered files until each is above {target}%",
1282 slots: {
1283 report: "coverage/coverage-summary.json",
1284 target: "80"
1285 }
11911286 },
11921287 "port-code-between-languages": {
11931288 title: "Port code to another language",
1194 teaches: "Say what to preserve, not just the target language. Naming the API or behavior that must stay the same gives Claude a contract to check the port against."
1289 teaches: "Say what to preserve, not just the target language. Naming the API or behavior that must stay the same gives Claude a contract to check the port against.",
1290 prompt: "port {source} to {target}, keeping the same {keep}",
1291 slots: {
1292 source: "this Python module",
1293 target: "Rust",
1294 keep: "public API and test behavior"
1295 }
11951296 },
11961297 "generate-docs-for-code": {
11971298 title: "Generate docs for undocumented code",
1198 teaches: "Name the scope and the format. Claude finds what's missing and matches the comment style already in the file, so the new docs read like the rest."
1299 teaches: "Name the scope and the format. Claude finds what's missing and matches the comment style already in the file, so the new docs read like the rest.",
1300 prompt: "find {scope} without {format} comments and add them, matching the style already used in the file",
1301 slots: {
1302 scope: "the public functions in src/auth/",
1303 format: "JSDoc"
1304 }
11991305 },
12001306 "migrate-a-pattern-across": {
12011307 title: "Migrate a pattern across the codebase",
1202 teaches: "Describe the old pattern and the new one. Asking Claude to identify every place first means the call sites are listed in the response, so you can check none were missed. For a migration across many files, run [/batch](/docs/en/commands). Claude splits the work into units for you to approve, then background subagents make the changes."
1308 teaches: "Describe the old pattern and the new one. Asking Claude to identify every place first means the call sites are listed in the response, so you can check none were missed. For a migration across many files, run [/batch](/docs/en/commands). Claude splits the work into units for you to approve, then background subagents make the changes.",
1309 prompt: "migrate everything from {from} to {to}: identify every place that needs to change, then make the changes",
1310 slots: {
1311 from: "the old logging API",
1312 to: "the structured logger"
1313 }
12031314 },
12041315 "optimize-against-a-measurable": {
12051316 title: "Optimize against a measurable target",
12061317 teaches: "Stating the metric and target gives Claude a clear definition of done.",
1207 next: "Set this as a `/goal` so Claude keeps measuring and iterating toward the number"
1318 next: "Set this as a `/goal` so Claude keeps measuring and iterating toward the number",
1319 prompt: "optimize {target} to bring {metric} from {current} down to under {goal}",
1320 slots: {
1321 target: "the search query",
1322 metric: "p95 latency",
1323 current: "2s",
1324 goal: "500ms"
1325 }
12081326 },
12091327 "fix-a-precise-visual": {
12101328 title: "Fix a precise visual bug",
12111329 teaches: "Precise visual feedback gets a precise fix. State the exact element, measurement, and viewport.",
1212 next: "Add a preview tool so Claude screenshots and verifies the fix itself"
1330 next: "Add a preview tool so Claude screenshots and verifies the fix itself",
1331 prompt: "the {element} extends {amount} beyond the {container} on {viewport}. fix it.",
1332 slots: {
1333 element: "login button",
1334 amount: "20px",
1335 container: "card border",
1336 viewport: "mobile"
1337 }
12131338 },
12141339 "review-your-changes-before": {
12151340 title: "Review your changes before you commit",
12161341 teaches: "Catch problems while they're still cheap to fix. Claude reads the changed files in full, not just the diff lines, so it spots issues a quick self-review misses.",
1217 next: "Run `/code-review` for the same check in one command"
1342 next: "Run `/code-review` for the same check in one command",
1343 prompt: "review my uncommitted changes and flag anything that looks risky before I commit"
12181344 },
12191345 "review-a-pull-request": {
12201346 title: "Review a pull request",
12211347 teaches: "Claude reviews with the whole codebase in context, not just the diff. It reads the changed code and what it calls, so it catches problems a diff-only review would miss.",
1222 next: "Run `/code-review <pr#>` in one command, or turn on Code Review for every PR"
1348 next: "Run `/code-review <pr#>` in one command, or turn on Code Review for every PR",
1349 prompt: "review PR #{pr} and summarize what changed, then list any concerns",
1350 slots: {
1351 pr: "247"
1352 }
12231353 },
12241354 "review-infrastructure-changes-before": {
12251355 title: "Review infrastructure changes before applying",
1226 teaches: "Plan output is dense and hard to scan. Pasting it gets you a plain-language summary of what's actually going to change before you apply it."
1356 teaches: "Plan output is dense and hard to scan. Pasting it gets you a plain-language summary of what's actually going to change before you apply it.",
1357 prompt: "here is my Terraform plan output. what is this going to do, and is anything here going to cause problems?"
12271358 },
12281359 "run-a-security-review": {
12291360 title: "Run a security review with a subagent",
12301361 teaches: "A [subagent](/docs/en/sub-agents) runs the audit in its own context window and reports back a summary, so a long security review doesn't fill up your main session. The built-in general-purpose subagent handles this without extra setup.",
1231 next: "Set up a dedicated security-review subagent your whole team can use"
1362 next: "Set up a dedicated security-review subagent your whole team can use",
1363 prompt: "use a subagent to review {path} for security issues and report what it finds",
1364 slots: {
1365 path: "src/api/"
1366 }
12321367 },
12331368 "review-content-before-sending": {
12341369 title: "Catch issues before formal review",
12351370 teaches: "Get a first pass before a human spends time on it. Name the concerns you want checked so the review is focused, then fix what it finds and send a cleaner draft.",
1236 next: "Capture your review checklist as a skill your whole team can run"
1371 next: "Capture your review checklist as a skill your whole team can run",
1372 prompt: "review {file} for {concerns} and list anything I should fix before it goes to {reviewer}",
1373 slots: {
1374 file: "launch-post.md",
1375 concerns: "unsupported claims, missing attributions, and brand-guideline issues",
1376 reviewer: "legal"
1377 }
12371378 },
12381379 "course-correct-a-wrong": {
12391380 title: "Course-correct a wrong approach",
12401381 teaches: "Name the constraint Claude missed, not just that it's wrong. A specific reason gives Claude a concrete constraint to satisfy on the retry, instead of guessing again.",
1241 next: "Press `Esc` twice to open the rewind menu and restore code and conversation so the retry starts clean"
1382 next: "Press `Esc` twice to open the rewind menu and restore code and conversation so the retry starts clean",
1383 prompt: "that is not right: {feedback}. try a different approach",
1384 slots: {
1385 feedback: "the function signature needs to stay backward-compatible"
1386 }
12421387 },
12431388 "narrow-the-scope-of": {
12441389 title: "Narrow the scope of a change",
1245 teaches: "When the direction is right but the change went too broad, ask Claude to keep part of it rather than rewinding everything. A stated boundary keeps a small fix from becoming a refactor."
1390 teaches: "When the direction is right but the change went too broad, ask Claude to keep part of it rather than rewinding everything. A stated boundary keeps a small fix from becoming a refactor.",
1391 prompt: "that is too much. keep only the changes to {scope} and undo your other edits",
1392 slots: {
1393 scope: "the validation logic in src/forms/"
1394 }
12461395 },
12471396 "turn-a-correction-into": {
12481397 title: "Turn a correction into a rule",
12491398 teaches: "A correction in chat isn't shared with your team. A rule in the project's [CLAUDE.md](/docs/en/memory) is shared once you commit it, and Claude reads it at the start of every session.",
1250 next: "Open `/memory` to review what Claude wrote"
1399 next: "Open `/memory` to review what Claude wrote",
1400 prompt: "you keep {mistake}. add a rule to CLAUDE.md so this stops happening",
1401 slots: {
1402 mistake: "using default exports when this project uses named exports"
1403 }
12511404 },
12521405 "resolve-merge-conflicts": {
12531406 title: "Resolve merge conflicts",
1254 teaches: "Say what state you want, not which markers to keep. Asking for the reasoning makes the merge reviewable instead of a black box."
1407 teaches: "Say what state you want, not which markers to keep. Asking for the reasoning makes the merge reviewable instead of a black box.",
1408 prompt: "resolve the merge conflicts in this branch and explain what you kept from each side"
12551409 },
12561410 "commit-with-a-generated": {
12571411 title: "Commit with a generated message",
1258 teaches: "Let Claude derive the message from the diff. It matches your repository's existing commit style."
1412 teaches: "Let Claude derive the message from the diff. It matches your repository's existing commit style.",
1413 prompt: "commit these changes with a message that summarizes what I did"
12591414 },
12601415 "open-a-pull-request": {
12611416 title: "Open a pull request from a ticket",
1262 teaches: "Skip the context switch between tracker, editor, and GitHub. One prompt reads the spec, makes the change, and opens the PR."
1417 teaches: "Skip the context switch between tracker, editor, and GitHub. One prompt reads the spec, makes the change, and opens the PR.",
1418 prompt: "find the {tracker} ticket about {topic} and open a PR that implements it",
1419 slots: {
1420 tracker: "Linear",
1421 topic: "the login timeout"
1422 }
12631423 },
12641424 "draft-release-notes-from": {
12651425 title: "Draft release notes from git history",
12661426 teaches: "Give two reference points and the structure you want. Claude reads the commit log between them and drafts a changelog you can edit.",
1267 next: "Save this as a `/changelog` skill"
1427 next: "Save this as a `/changelog` skill",
1428 prompt: "compare {from} to {to} and draft release notes grouped by feature, fix, and breaking change",
1429 slots: {
1430 from: "v2.3.0",
1431 to: "v2.4.0"
1432 }
12681433 },
12691434 "write-a-ci-workflow": {
12701435 title: "Write a CI workflow",
1271 teaches: "Describe when it should run and what it should do; the YAML is generated for you, matched to your project's build and test commands."
1436 teaches: "Describe when it should run and what it should do; the YAML is generated for you, matched to your project's build and test commands.",
1437 prompt: "write a GitHub Actions workflow that {steps} on every push to {branch}",
1438 slots: {
1439 steps: "runs the tests and deploys to staging",
1440 branch: "main"
1441 }
12721442 },
12731443 "find-and-fix-a": {
12741444 title: "Find and fix a failing test",
1275 teaches: "Describe the symptom; you don't need to know which file is broken. Claude runs the test to see the failure, traces it into source, and fixes it."
1445 teaches: "Describe the symptom; you don't need to know which file is broken. Claude runs the test to see the failure, traces it into source, and fixes it.",
1446 prompt: "the {test} test is failing, find out why and fix it",
1447 slots: {
1448 test: "UserAuth"
1449 }
12761450 },
12771451 "investigate-a-reported-error": {
12781452 title: "Investigate a reported error",
12791453 teaches: "Describe the symptom and location; Claude reads the relevant code path and traces likely causes. Paste stack traces or logs if you have them.",
1280 next: "Put a deeplink in your runbook that opens Claude with this prompt pre-filled"
1454 next: "Put a deeplink in your runbook that opens Claude with this prompt pre-filled",
1455 prompt: "users are seeing {symptom} on {where}. investigate and tell me what is going on",
1456 slots: {
1457 symptom: "500 errors",
1458 where: "/api/settings"
1459 }
12811460 },
12821461 "fix-a-build-error": {
12831462 title: "Fix a build error at the root",
1284 teaches: "Asking for root cause and verification prevents surface-level patches that suppress the error without fixing it."
1463 teaches: "Asking for root cause and verification prevents surface-level patches that suppress the error without fixing it.",
1464 prompt: "here is a build error. fix the root cause and verify the build succeeds"
12851465 },
12861466 "investigate-a-production-incident": {
12871467 title: "Investigate a production incident",
12881468 teaches: "List the evidence sources to correlate, not the steps to take. Claude reads logs, git history, and config together to narrow the cause.",
1289 next: "Connect Sentry or your log store via MCP"
1469 next: "Connect Sentry or your log store via MCP",
1470 prompt: "{symptom}. check the logs, recent deploys, and config changes, then tell me the most likely cause",
1471 slots: {
1472 symptom: "the checkout endpoint started returning 500s an hour ago"
1473 }
12901474 },
12911475 "query-logs-in-plain": {
12921476 title: "Query logs in plain English",
1293 teaches: "Ask the question instead of writing the SQL. Claude builds the query, runs it against your connected logs, and shows both the query and the result so you can check what ran."
1477 teaches: "Ask the question instead of writing the SQL. Claude builds the query, runs it against your connected logs, and shows both the query and the result so you can check what ran.",
1478 prompt: "show me all {events} for {scope} over {timeframe}. write the query, run it, and tell me what stands out",
1479 slots: {
1480 events: "failed logins",
1481 scope: "the auth service",
1482 timeframe: "the past 24 hours"
1483 }
12941484 },
12951485 "diagnose-from-a-console": {
12961486 title: "Diagnose from a console screenshot",
1297 teaches: "Cloud consoles show you the problem but not the commands to fix it. Claude reads the screenshot and translates the dashboard into the kubectl, gcloud, or aws commands to run."
1487 teaches: "Cloud consoles show you the problem but not the commands to fix it. Claude reads the screenshot and translates the dashboard into the kubectl, gcloud, or aws commands to run.",
1488 prompt: "here is a screenshot of {console}. walk me through why {resource} is failing and give me the exact commands to fix it",
1489 slots: {
1490 console: "the GCP Kubernetes dashboard",
1491 resource: "this pod"
1492 }
12981493 },
12991494 "analyze-a-data-file": {
13001495 title: "Analyze a data file",
13011496 teaches: "A one-off question doesn't need a one-off script. Point at a file in your project folder and Claude reads it directly, finds the patterns, and writes the output where you ask.",
1302 next: "Connect the data source via MCP instead of exporting files"
1497 next: "Connect the data source via MCP instead of exporting files",
1498 prompt: "read {file}, summarize the key patterns, and write the results to {output}",
1499 slots: {
1500 file: "@reports/q1-signups.csv",
1501 output: "an HTML page with charts, then open it in my browser"
1502 }
13031503 },
13041504 "generate-variations-from-performance": {
13051505 title: "Generate variations from performance data",
13061506 teaches: "State the constraint at the start so generation stays within the limit. Claude reads the metrics, picks what to replace, and produces alternatives that fit.",
1307 next: "Connect the ad platform via MCP instead of exporting a file"
1507 next: "Connect the ad platform via MCP instead of exporting a file",
1508 prompt: "read {file}, find the underperforming {items}, and generate {n} new variations that stay under {limit} characters",
1509 slots: {
1510 file: "@ads-performance.csv",
1511 items: "headlines",
1512 n: "20",
1513 limit: "90"
1514 }
13081515 },
13091516 "turn-a-recurring-task": {
13101517 title: "Turn a recurring task into a skill",
1311 teaches: "Name the steps once; reuse them as a command. Claude writes a [skill](/docs/en/skills) anyone on your team can run."
1518 teaches: "Name the steps once; reuse them as a command. Claude writes a [skill](/docs/en/skills) anyone on your team can run.",
1519 prompt: "create a /{name} skill for this project that {steps}",
1520 slots: {
1521 name: "ship",
1522 steps: "runs the linter and tests, then drafts a commit message"
1523 }
13121524 },
13131525 "add-a-hook-for": {
13141526 title: "Add a hook for repeat behavior",
1315 teaches: "Hooks make a behavior automatic instead of something you have to remember to ask for. Describe the trigger and action and Claude writes the [hook](/docs/en/hooks) configuration."
1527 teaches: "Hooks make a behavior automatic instead of something you have to remember to ask for. Describe the trigger and action and Claude writes the [hook](/docs/en/hooks) configuration.",
1528 prompt: "write a hook that {action} after every {event}",
1529 slots: {
1530 action: "runs prettier",
1531 event: "edit to a .ts or .tsx file"
1532 }
13161533 },
13171534 "connect-a-tool-with": {
13181535 title: "Connect a tool with MCP",
1319 teaches: "Connect the source once instead of pasting data every session. After [MCP](/docs/en/mcp) setup, Claude reads from the tool directly when you ask about it."
1536 teaches: "Connect the source once instead of pasting data every session. After [MCP](/docs/en/mcp) setup, Claude reads from the tool directly when you ask about it.",
1537 prompt: "set up the {server} MCP server so you can read my {data} directly",
1538 slots: {
1539 server: "Sentry",
1540 data: "error reports"
1541 }
13201542 },
13211543 "capture-what-to-remember": {
13221544 title: "Capture what to remember for next time",
1323 teaches: "Ask before you forget. Claude knows what it had to figure out this session and proposes [CLAUDE.md](/docs/en/memory) entries so the next session starts with that context."
1545 teaches: "Ask before you forget. Claude knows what it had to figure out this session and proposes [CLAUDE.md](/docs/en/memory) entries so the next session starts with that context.",
1546 prompt: "summarize what we did this session and suggest what to add to CLAUDE.md"
13241547 }
13251548};
13261549
No line in this hunk matches that.