data/opus-12.jsonl

{"prompt": "would you mind auditing the checkout flow against wcag 2.2 AA and telling me what fails? no fixes yet, i need the list for a stakeholder meeting", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "the date picker is a div soup with no keyboard support at all. rebuild it as a proper combobox+grid with roving tabindex", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "aria-label on the icon-only buttons in the toolbar, there are six of them", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "boundary", "lang": "en"}
{"prompt": "screen reader announces the row count wrong after filtering — says 40 when 3 are shown. live region timing i think but i'm guessing", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "our components each hand-roll their focus trap. one shared hook, identical behavior including the shift-tab wrap", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "put together the accessibility plan for the next two quarters — what we fix, what we test automatically, and how we stop regressions", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "write our accessibility guidelines for engineers, with the five patterns we get wrong most and how to do them right", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "keyboard shortcut for skip-to-content plus the visible skip link, and it needs to actually move focus not just scroll", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "python: incremental loader for the vendor sftp drops — new files only, checksum verified, and it should quarantine anything that fails schema validation instead of failing the whole run", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "could you explain what the airflow sensor is actually waiting on in the vendor dag? i inherited it and the poke interval seems wrong", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "the schedule_interval is '@daily' but the vendor drops at 23:40 local so we always process yesterday's file", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "our 14 dags each define their own default_args with slightly different retries and owners. one factory, same effective config per dag — list any that change", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "one file out of about 300 a day produces rows with the columns shifted right by one. the file looks fine when i open it", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "data contract doc for the vendor feed — every column, type, nullability, and what we do when they break it (they will)", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "please plan how we move from airflow to dagster, or decide not to. 14 dags, heavy on sensors, and two people who know airflow", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "swift command line tool that watches a directory and re-signs any new .app bundle it finds, with a launchd plist to keep it running", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "the tool prints progress with \\r which looks broken in xcode's console", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "our helper tool and the main app both parse the same plist format with separate code. share it via a small package, identical parsing on the fixture set", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "codesign fails in CI with 'resource fork, Finder information, or similar detritus not allowed' and works on my machine", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "menu bar app with a settings window, launch at login, and a graceful path for when the accessibility permission is denied", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "review the entitlements we request and tell me which ones we don't actually need", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "readme for the helper tool: install, the permissions it needs and why, and how to verify it's running", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "map out the sandboxing work for the mac app so we can ship on the app store. what breaks, what needs a temporary exception, what we'd have to drop", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "xpc service for the privileged operations, with a tight allowlist of what the app can ask it to do", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "onward", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "vague-eval", "lang": "en"}
{"prompt": "pasting what the a11y consultant sent, please work through it:\n\nBLOCKERS\n1. Modal dialogs: focus is not moved into the dialog on open; background content is not\n   hidden from AT (no aria-hidden / inert on the page container).\n2. Data table sort buttons: no aria-sort, and the sort state is only conveyed by an icon.\n3. Form errors: announced via a toast that disappears in 4s; not associated with the\n   fields via aria-describedby.\n4. Custom select: uses role=\"listbox\" but options are divs without role=\"option\".\n5. Contrast: secondary button text #8A8F98 on #F4F5F7 = 2.9:1.\n\nMINOR\n6. Decorative icons missing aria-hidden.\n7. Page titles identical across routes.\n\nWe'd want blockers resolved before the audit report is finalized.", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "pasted-context", "lang": "en"}
{"prompt": "the secondary button text color needs to hit 4.5:1, pick the nearest token that does", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "boundary", "lang": "en"}
{"prompt": "our page titles are all 'Acme' — set them per route from the route metadata", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "plan the design token overhaul so contrast is guaranteed by construction, not checked after the fact", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "form field component that wires label, description, and error to the input with the right aria attributes, so we stop getting this wrong per-form", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "explain how our current toast-based error announcement works and why AT users miss it", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "axe checks in CI on the 10 main routes, failing the build on new violations but not on the existing backlog", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "our three select implementations should be one. keep the native-ish behavior of the newest, and every existing usage must still work", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "accessibility conformance report (VPAT-ish) draft based on the audit findings — factual, with the known gaps listed honestly", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "voiceover reads the table header twice for every cell in safari but not in chrome", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "spark structured streaming job from kafka to iceberg, exactly-once with checkpointing, and schema evolution handled without a restart", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.9, "slice": "core", "lang": "en"}
{"prompt": "our batch and streaming jobs compute the same aggregates with duplicated logic. share the transformation, and prove the outputs match on a day of data", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "the checkpoint location is in /tmp", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "would you write up the lakehouse layout for us — bronze/silver/gold, partitioning per table, and the compaction schedule", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "streaming job falls behind by hours overnight then catches up by 9am. the input rate is flat", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "read the iceberg table properties we set and tell me whether small file compaction is actually running", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "figure out the plan for backfilling three years of history into the new lakehouse tables without competing with the live stream", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "data quality dashboard: freshness per table, row count trend, and the failed expectation list from the last run", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "design the lineage story and then build the collector that reads it from the spark listener", "purpose": "planning", "secondary": "backendImpl", "mixed": true, "difficulty": 0.8, "slice": "mixed", "lang": "en"}
{"prompt": "find out why the freshness metric is wrong and add a note to the data catalog entry", "purpose": "debugging", "secondary": "writing", "mixed": true, "difficulty": 0.6, "slice": "mixed", "lang": "en"}
{"prompt": "picking this back up, whats left", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "vague-eval", "lang": "en"}
{"prompt": "we're getting asked for an audit trail by two enterprise prospects and i want to do it once, properly. every mutation, who and what changed, immutable, exportable, queryable by resource and actor, and retained for 7 years without bloating the primary db. write the design", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "audit event capture at the orm layer so no code path can skip it, with the diff stored as a compact json patch", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "audit log viewer: filters for actor, resource type, action and date, a diff view per entry, and csv export of the filtered set", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "the audit table has no index on (resource_type, resource_id)", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "we log audit events from 30 call sites with hand-built payloads. move to the orm hook and remove the manual calls, same events recorded", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "audit entries have a null actor for about 4% of events and i can't tell which code path produces those", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "is the audit log actually append-only, or can an admin edit it through some path? check the grants and the code", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "compliance-facing doc describing our audit logging: coverage, retention, integrity, and access controls", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "audit export to a customer's s3 bucket nightly, in a documented format, with a manifest", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "export settings page: bucket, role arn, a test connection button, and the last export status", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "the export filename uses the local date so we get gaps and duplicates around midnight utc", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "plan the archival tier for audit data — hot for 90 days in postgres, then s3 with a query path that isn't a lie", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "queries against audit data older than 90 days return empty instead of erroring, so people think there's no data", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "walk me through how a customer's audit export could include another customer's events, if at all", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "docs for the audit export format — one page, the json schema, and a sample file", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "boundary", "lang": "en"}
{"prompt": "extract the diffing code used by audit, version history and the deploy view into one utility. same diffs everywhere", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "por favor, planifica cómo vamos a migrar los reportes a la nueva API sin romper los clientes actuales, con fases", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "es"}
{"prompt": "report scheduling: cron per report, delivery by email or s3, timezone per recipient, and skip if the report is empty", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "schedule editor with a human-readable summary under the cron input ('every weekday at 8am in Europe/Berlin')", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "the cron input accepts 6 fields but our runner only reads 5", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "scheduled reports fire twice for recipients in timezones with a half-hour offset", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "pdf rendering for reports, headless chrome, with page breaks that don't split table rows and a header on every page", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "the pdf uses a font that isn't embedded so cyrillic renders as boxes", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "check our pdf pipeline for whether a report can include data the recipient isn't allowed to see", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "help article on scheduled reports: setting one up, the timezone rules, and why an empty report doesn't send", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "yep", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "vague-eval", "lang": "en"}
{"prompt": "from the ops channel, please sort out the top two:\n\n> @ravi: heads up, three things piling up\n> 1. the nightly `vacuum_analyze` cron has been failing for 11 days, exit 1, no output captured\n> 2. our terraform state bucket has no versioning on it (found while doing the disaster recovery doc)\n> 3. staging db is a 6 month old dump, people keep testing against unrealistic data\n>\n> also unrelated but the grafana admin password is in the wiki 🙃", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "pasted-context", "lang": "en"}
{"prompt": "state bucket needs versioning and a lifecycle rule, plus block public access explicitly", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "figure out our disaster recovery posture and write the plan — rpo, rto, what we actually test, and where we're lying to ourselves", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "automated staging refresh from a scrubbed prod dump, weekly, with pii replaced deterministically so referential integrity holds", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "the scrubbing script and the seed script both fake user data with different rules. one faker module, same shapes", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "cron reports success in our monitoring but the actual command fails, so we didn't notice for 11 days", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "review our backup setup end to end and tell me whether we could actually restore, with the steps someone would follow", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "restore runbook, tested — write it as a checklist with the exact commands and expected output at each step", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "internal status board for scheduled jobs: last run, duration, exit code, and a red row if it hasn't run in 2x its interval", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "cron wrapper that captures stdout/stderr, reports the exit code to our monitoring, and dead-man-switches if it doesn't run", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "our crontab has 22 entries with inline bash, half with no logging. move them to a scheduled job definition file, same schedules and commands", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "the vacuum cron runs at 2am utc which overlaps the backup window", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "plan how we get secrets out of the wiki and into something real, including rotating everything that's been sitting there", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "grafana's admin password, the ci token and two api keys have been in the wiki. rotate them and tell me what depends on each", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "explain what our current monitoring would and wouldn't catch if the primary database went read-only", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "write the security incident procedure for a leaked credential — who does what, in what order, and the comms", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "one more thing then we're done", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "vague-eval", "lang": "en"}
{"prompt": "i'd like a design for the multi-step approval workflow the enterprise customers keep asking for. arbitrary approver chains, conditional steps based on amount, delegation when someone's on holiday, and a full audit of who approved what. this feels like it wants to be a state machine but i want your take before we build", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.9, "slice": "core", "lang": "en"}
{"prompt": "approval workflow engine: definitions in json, instances tracked per request, and the transitions validated server side", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "approval chain visualizer — steps as a horizontal flow, current step highlighted, completed ones with the approver and timestamp", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "the approve button is enabled for people who aren't the current approver", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "the approval state lives in three columns plus a json blob and they can disagree. one representation, and migrate the in-flight instances safely", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "requests occasionally get stuck with no current approver and no way forward except a db edit", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "read the delegation code and tell me if a delegate can approve something the delegator couldn't", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "admin documentation for building approval chains, with three worked examples and the gotchas about conditional steps", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "reminder emails for pending approvals — after 24h, then daily, stopping when it moves on, and never on weekends", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "pending approvals inbox with bulk approve for the low-risk category and a required comment for rejections", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "reminders keep sending after the request is approved", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "the weekend check uses the server's timezone not the approver's", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "boundary", "lang": "en"}
{"prompt": "sketch how we'd let customers define their own conditions without giving them arbitrary code execution", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "expression evaluator for approval conditions — a small safe language, no loops, typed against the request fields", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "condition builder ui: field, operator, value rows with and/or, and a live 'this would match' indicator against a sample request", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "explain how our expression evaluator handles a missing field — null, error, or false? i need to know for the docs", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "reference page for the condition expression language: types, operators, functions, and the evaluation rules for nulls", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "our validation of conditions happens client side only, which is obviously not enough. move the check server side and keep the client one for ux", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "boundary", "lang": "en"}
{"prompt": "les conditions avec des montants négatifs passent alors qu'elles ne devraient pas. je ne comprends pas pourquoi", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "fr"}
{"prompt": "fine, next", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "vague-eval", "lang": "en"}
{"prompt": "could you plan the internationalization work for the product? 40k strings-ish, plurals, rtl for arabic, and the dates and numbers are formatted ad hoc everywhere. i want phases and an idea of what the translation workflow looks like", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "extraction and message catalog pipeline — pull strings from the code, push to the translation service, pull back and typecheck the keys", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "rtl layout support: logical properties throughout, mirrored icons where appropriate, and the sidebar flipping correctly", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "the language switcher lists 'Portuguese' twice, one is pt-BR", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "we concatenate strings for sentences in 60 places which is untranslatable. move to full sentences with placeholders, same rendered english", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "german text overflows the buttons on the dashboard and the labels clip mid-word", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "explain how our plural handling works today, i think it's just a ternary on n !== 1 which won't fly for polish", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "localization guide for engineers: how to add a string, plural rules, what never to concatenate, and how to test a pseudo-locale", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "currency and number formatting through one helper that respects the user's locale, replacing the 30 ad hoc format calls", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "pseudo-locale mode in dev that expands strings 40% and brackets them, toggled by a query param", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "dates show as '2026/7/30' for en-GB users", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "plan how we handle user-generated content in mixed languages — do we detect, do we translate, what do we show", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "translated emails render with the english subject line for two of the six locales", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "language picker that remembers the choice per account, falls back to the browser preference, and doesn't flash english first", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "tell me which of our strings are still hardcoded in the templates, and roughly how many", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "translator notes for the 200 strings with ambiguous context — the ones where 'Post' could be a verb or a noun", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "design the offline mode for the web app and then implement the service worker caching layer", "purpose": "planning", "secondary": "frontendImpl", "mixed": true, "difficulty": 0.8, "slice": "mixed", "lang": "en"}
{"prompt": "look over the i18n helper and clean up the api while keeping the same output", "purpose": "review", "secondary": "refactor", "mixed": true, "difficulty": 0.5, "slice": "mixed", "lang": "en"}
{"prompt": "carry on", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "vague-eval", "lang": "en"}
{"prompt": "here's the whole failing CI output, i've read it four times and i'm none the wiser:\n\nRun make integration\ndocker compose -f compose.ci.yml up -d --wait\n[+] Running 4/5\n ✔ Container ci-postgres-1  Healthy                        8.1s\n ✔ Container ci-redis-1     Healthy                        2.0s\n ✔ Container ci-minio-1     Healthy                        3.4s\n ✘ Container ci-api-1       Error                         31.2s\ndependency failed to start: container ci-api-1 exited (1)\n\ndocker compose logs api:\nci-api-1  | 2026/07/30 06:11:02 connecting to postgres host=postgres port=5432 db=acme\nci-api-1  | 2026/07/30 06:11:02 migrate: dirty database version 214, fix and force version\nci-api-1  | 2026/07/30 06:11:02 startup failed: migrate up: Dirty database version 214\n\nmake: *** [Makefile:41: integration] Error 1\n\nthe postgres container is fresh every run so how is it dirty", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "pasted-context", "lang": "en"}
{"prompt": "migration runner should fail loudly on a dirty state at startup instead of after partially connecting, and never leave it dirty on a failed apply", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "the compose healthcheck for api uses curl which isn't in the image", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.2, "slice": "core", "lang": "en"}
{"prompt": "plan how we make local dev and CI use the same compose setup so 'works on my machine' stops being a thing", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "we have compose.yml, compose.ci.yml and compose.local.yml with 80% overlap and three different postgres versions. consolidate with overrides, same effective services per environment", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "explain what the --wait flag actually waits for and whether our healthchecks mean what we think", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "boundary", "lang": "en"}
{"prompt": "developer setup doc that works from a clean machine, including the two things people always get stuck on", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "seeded dev data that's actually representative — 5 orgs of different sizes, a big one with 50k records, and edge cases like unicode names", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "dev toolbar in the app showing the current user, org, feature flags and a way to switch, only in non-prod", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "the dev toolbar renders in production if you set the query param", "purpose": "quickFix", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "core", "lang": "en"}
{"prompt": "tests pass on linux and fail on macos with a file path assertion", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.5, "slice": "core", "lang": "en"}
{"prompt": "review the makefile and tell me which targets are broken or reference things that don't exist", "purpose": "review", "secondary": null, "mixed": false, "difficulty": 0.4, "slice": "core", "lang": "en"}
{"prompt": "write the architecture overview for new joiners — the six services, what talks to what, and where to start reading", "purpose": "writing", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "map out the plan for a proper preview environment per PR — database, seeded data, and a url in the PR comment", "purpose": "planning", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "preview environment provisioner: namespace per PR, a database branch, teardown on merge or after 3 days idle", "purpose": "backendImpl", "secondary": null, "mixed": false, "difficulty": 0.8, "slice": "core", "lang": "en"}
{"prompt": "our test helpers reach into private methods in 40 places which makes any refactor a nightmare. rework them to use public seams, same coverage", "purpose": "refactor", "secondary": null, "mixed": false, "difficulty": 0.7, "slice": "core", "lang": "en"}
{"prompt": "preview envs leak — we have 60 namespaces for 12 open PRs", "purpose": "debugging", "secondary": null, "mixed": false, "difficulty": 0.6, "slice": "core", "lang": "en"}
{"prompt": "figure out the design for our staging data privacy story, then implement the scrubber", "purpose": "planning", "secondary": "backendImpl", "mixed": true, "difficulty": 0.7, "slice": "mixed", "lang": "en"}
{"prompt": "just finish the last bit please", "purpose": "frontendImpl", "secondary": null, "mixed": false, "difficulty": 0.3, "slice": "vague-eval", "lang": "en"}