promptstudio
promptstudio's Prompts
Claude Code Hooks Config Writer for Team Repos: settings.json PreToolUse and PostToolUse Matchers, Exit Code 2 Blocking, stdin JSON Parsing with jq, Auto Format After Edits, Stop Hook Loop Guard, and a Hook Test Plan
Write Claude Code hooks a team can commit: the right settings file scope, PreToolUse guards that block destructive shell commands and edits to protected paths with exit code 2 and a reason Claude can read, PostToolUse formatting on edited files, a Stop hook that checks work without looping forever, and a plan to test each hook before trusting it.
Act as a developer productivity engineer who maintains Claude Code setups for a team, writes hooks as small shell or Python scripts under .claude/hooks, and treats every hook as code that runs with the developer's own permissions. Inputs: - Repo stack, package manager, formatter, linter, and test or typecheck commands: [RepoStack] - Commands or actions Claude must never run without asking: [BlockedActions] - Files or folders Claude must not edit: [ProtectedPaths] - What should happen automatically after edits or at the end of a task: [Automations] - Scope: personal only, shared with the team, or both: [Scope] - Output format: [Format] Generate: 1. Settings file choice from Scope: user settings in the home .claude folder, project settings in .claude/settings.json committed to git, or .claude/settings.local.json for personal overrides that stay out of git. 2. A hooks block in valid JSON: hook events as keys (PreToolUse, PostToolUse, Stop, and Notification or UserPromptSubmit only if Automations needs them), each with a matcher for tool names such as Bash or Edit|Write|MultiEdit, and command hooks pointing at scripts via the CLAUDE_PROJECT_DIR environment variable. 3. A PreToolUse guard for BlockedActions: read the JSON on stdin with jq, pull tool_input.command, match the dangerous patterns, print a reason to stderr, and exit 2 so the call is blocked and Claude sees why. Exit 0 lets the call through. 4. A path guard for ProtectedPaths on edit tools: read tool_input.file_path, block matches with exit 2 and a clear message. 5. PostToolUse automations from Automations and RepoStack: format only the edited file, keep output quiet, set a timeout, and never fail the session because a formatter is missing. 6. A Stop hook when Automations asks for a final check: run the typecheck or tests, exit 2 with the failures on stderr so Claude keeps working, and read stop_hook_active from stdin so it does not loop forever. 7. Each script in full with a shebang, set options, and a comment on what it reads and returns. Remind the team to make scripts executable. 8. A test plan: pipe sample JSON into each script by hand and check the exit code, then open the hooks menu in Claude Code to confirm they loaded, then try one blocked command in a scratch branch. 9. A security note: hooks run arbitrary commands with your account's access, so review hook changes in pull requests like any other code. Constraints: - Only use hook events, fields, and exit code behavior you are sure of; mark anything else "confirm in the current Claude Code hooks docs". - Quote paths that may contain spaces. No em dashes.
Meta Business Portfolio Access Audit and Agency Offboarding Plan: Asset Ownership Map for Pages, Instagram, Ad Accounts, Datasets and Catalogs, Full Control vs Partial Access Roles, Partner Removal, Two Factor Rules, and a Quarterly Review
Take back control of a brand's Facebook and Instagram setup before or after an agency or employee leaves: map which business portfolio owns each Page, Instagram account, ad account, dataset, and catalog, fix assets that sit in someone else's portfolio, trim people to the right role, remove partner access in the right order, require two factor authentication, and set a quarterly access review.
Act as a paid social operations lead who has audited dozens of Meta business portfolios (formerly Business Manager) for brands switching agencies, and has seen a brand lose years of pixel history because the agency owned the ad account and dataset. Inputs: - Assets as listed in Business settings: Pages, Instagram accounts, ad accounts, datasets or pixels, catalogs, domains, apps, with which portfolio owns each: [AssetInventory] - People with access and their current access level: [PeopleList] - Partner businesses (agencies, freelancers, tools) and the assets shared with them: [Partners] - What triggered the audit (agency change, employee exit, hacked account, routine review): [Trigger] - Ad billing setup: who owns the payment method on each ad account: [Billing] - Output format: [Format] Generate: 1. An asset ownership map from AssetInventory: asset, owning portfolio, shared with, and status (owned by us, owned by partner, personal account only). Explain that a partner can be removed from an asset you own, but an asset owned in the partner's portfolio stays with them unless they move or release it. 2. A rescue plan for partner owned assets: Pages and Instagram accounts the partner can transfer or release, a new brand owned ad account if the old one cannot move, and a new or shared dataset plan so conversion history is not lost silently. Mark exact transfer steps "check current Meta help". 3. A people review from PeopleList: at least two trusted employees with full control, everyone else on partial access with only the tasks they need, removal of people who left, and no personal accounts of former agency staff. 4. Partner access order from Partners and Trigger: first confirm the replacement has access, then remove the outgoing partner from each asset, then remove the partner business itself, then check Page and Instagram roles outside the portfolio. 5. Security settings: require two factor authentication for everyone in the portfolio, confirm the business email and admin contact, check the portfolio's security or alerts area for warnings, and verify the domain if not done. 6. Billing check from Billing: who owns each payment method, removing agency cards, adding the brand's own method before campaigns restart, and invoices or tax details. 7. A change log template: date, person, asset, action, before and after access. 8. A quarterly review routine: export people and partner lists, compare against staff and vendor lists, and confirm two full control admins. Constraints: - Meta renames menus often: describe the setting and mark "find in Business settings" instead of inventing exact paths. - Never recommend sharing login passwords or adding anyone to full control as a shortcut. No em dashes.
Journal Correction Request Writer for Published Articles: Error Inventory, Corrigendum vs Erratum vs Retraction Triage Using COPE Guidance, Impact on Conclusions, Correction Notice Text, Coauthor Approval, and an Editor Email
Fix a mistake in a published paper the right way: list each error with its exact location, decide whether it is an author corrigendum, a publisher erratum, or something more serious that needs the editor's judgment, state plainly whether the conclusions change, draft the correction notice text, collect coauthor approval, and write the email to the handling editor.
Act as a research integrity officer and journal editorial board member who handles post publication corrections, follows COPE guidance, and knows editors respond faster when an author arrives with a precise error list and draft notice text. Inputs: - Article citation, DOI, journal, and publisher: [ArticleDetails] - Each error found: where it is (section, page, table, figure, supplement), what it says, and what it should say: [ErrorList] - How each error happened (author mistake in the submitted file, introduced during typesetting, data or code problem): [ErrorOrigin] - Whether results, statistics, or conclusions change, with recalculated values: [ImpactCheck] - Author list and who has already agreed to the correction: [Authors] - Output format: [Format] Generate: 1. An error inventory table from ErrorList: number, location, published text, corrected text, and type (typo, wrong value, mislabeled figure, missing funding or conflict statement, affiliation, reference). 2. Triage for each item using ErrorOrigin: corrigendum when the error came from the authors, erratum when it was introduced by the publisher during production, and "editor decision" when ImpactCheck shows a conclusion changes, a key result no longer holds, or data reliability is in doubt, where an expression of concern, retraction and replacement, or retraction may be considered. Explain each term in one line and let the editor make the call. 3. An impact statement per item and overall: state clearly whether conclusions change, using the recalculated values in ImpactCheck. Never soften a change that alters a finding. 4. Draft correction notice text in journal style: the article reference, each correction quoted as published and as corrected, and a closing line on whether conclusions are affected. Keep it factual, with no apology paragraph. 5. Supporting files to send: corrected table or figure files, a marked up copy, recalculation code or output when values changed. 6. A coauthor approval checklist from Authors: who must sign off, a plain message to send them, and a note that unresolved disagreement should go to the editor rather than be hidden. 7. An email to the handling editor or the journal's corrections contact: subject line, what was found and how, the triage, the draft notice attached, and an offer to answer questions. 8. Follow up items: check the notice links to the original article once published, update the version in institutional and preprint repositories with a link to the correction, and correct any derived data or code release. Constraints: - Do not invent page numbers, values, or journal policy; mark gaps [CHECK PUBLISHED PDF] or [CHECK JOURNAL POLICY]. - This is a drafting aid, not a misconduct determination. No em dashes.
Ink Interactive Fiction Script Writer for inkle's Ink and Inky: Knot and Stitch Scene Map, Once Only and Sticky Choices, Gathers, VAR State Tracking, Conditional Text and Choices, Diverts to END, and a Playtest Path Table
Turn a branching story outline into a working Ink script you can paste into Inky: a knot and stitch map, once only and sticky choices that never leave the player stuck, gathers that fold branches back together, VAR variables for trust and items, conditional lines and choices, clean diverts to END, and a table of paths to playtest before you export to inkjs or Unity.
Act as a narrative designer who writes branching dialogue in inkle's Ink language, tests every scene in the Inky editor, and has shipped Ink stories through inkjs on the web and the Ink Unity integration. Inputs: - Story premise, tone, and point of view: [Premise] - Scene outline in order, with the decisions the player makes: [SceneOutline] - State to track (relationships, items, clues, counters): [StateList] - Endings you want and what unlocks each: [Endings] - Target runtime (inkjs web page, Unity, or Inky only) and any tags your engine reads: [Runtime] - Output format: [Format] Generate: 1. A scene map from SceneOutline: one knot per scene with a snake_case name, stitches for sub locations or beats, and a line for each divert between them. Flag any scene with no way out. 2. A variable table from StateList: each VAR with its starting value and type, where it changes, and where it is read. Use temp variables only for values inside one knot. 3. The script in valid Ink: knots with triple equals signs, stitches with one equals sign, choices with an asterisk when they should disappear after use and a plus sign when they must stay available, square bracket text so the choice wording is not repeated in the output when that reads better, and labeled gathers so dialogue loops back cleanly. 4. State changes with tilde lines, conditional choices that show only when a variable allows them, and inline conditional text for lines that change with state. 5. A multi line conditional block that routes to each ending in Endings, and every path ending in a divert to END or to another knot. No path should run out of choices. 6. Tags for Runtime only when the engine reads them (speaker, mood, audio cue), written with a hash mark at line end. 7. A playtest path table: each path as a list of choices, the variables at the end, the ending reached, and what to check in Inky (no dead ends, no repeated choice text, no orphan knots). 8. Export notes: compile with inklecate or Inky's export to JSON, and load the JSON in inkjs or the Unity integration according to Runtime. Constraints: - Every line of code must compile in current Ink. Do not invent functions or syntax; if unsure, keep it simple. - Keep prose lines short enough to read on a phone screen. No em dashes.
Sales Tax Economic Nexus Tracker for Online Sellers After Wayfair: State Threshold Table, Measurement Periods, Marketplace Facilitator Sales Treatment, Registration Order, Voluntary Disclosure Flags, and a Monthly Monitoring Routine
Turn an online seller's sales by state into a nexus action plan: compare each state's economic nexus threshold and measurement period from your own research, decide whether marketplace sales count toward the threshold, rank which states to register in first, flag past exposure that may need a voluntary disclosure agreement, and set up a monthly check so you register before collecting, not after.
Act as a sales tax compliance analyst for ecommerce brands who has run economic nexus studies since the 2018 South Dakota v. Wayfair decision, and knows that most mistakes come from counting the wrong sales against the threshold or collecting tax before registering. Inputs: - Sales by state for the last two calendar years and the current year to date, split into direct site sales and marketplace sales (Amazon, Etsy, Walmart): [SalesByState] - Transaction counts by state if you have them: [TransactionCounts] - State thresholds, measurement periods, and whether marketplace sales count, copied from each state revenue department site with the date checked: [StateThresholds] - Where you have physical presence today (inventory, employees, office, trade shows): [PhysicalPresence] - Products sold and anything that may be exempt or taxed differently (clothing, food, digital goods): [ProductTaxability] - Output format: [Format] Generate: 1. A threshold table using only StateThresholds: state, dollar threshold, transaction threshold if any, whether it is gross, retail, or taxable sales, measurement period (prior calendar year, current year, or rolling 12 months), and whether marketplace sales count. Any state without a row is marked "NOT RESEARCHED" rather than filled in from memory. 2. A nexus test for each state from SalesByState and TransactionCounts: the sales counted under that state's rule, the result (over, near at 80 percent or more, under), and the date the threshold was crossed if it was. 3. Marketplace facilitator treatment: the marketplace collects and remits on its own sales in most states, but those sales may still count toward your threshold. Show the test both ways where StateThresholds is unclear and mark "confirm with the state". 4. Physical presence nexus from PhysicalPresence, including inventory stored in a third party warehouse, listed separately because no threshold applies. 5. A registration order: states already over first, ordered by crossing date and sales volume, then states near the line. Note that you should not collect tax from customers in a state until you hold a permit there. 6. Past exposure flags: states crossed more than one filing period ago where you did not collect. Explain a voluntary disclosure agreement in plain words and mark these for a CPA or sales tax advisor. 7. ProductTaxability notes per registered state, marked "verify taxability in each state" for anything not tax at the standard rate. 8. After registration: expected filing frequency is assigned by the state, local rate and sourcing questions for home rule states, and a calendar for returns. 9. A monthly monitoring routine: export sales by ship to state, update the table, and alert at 80 percent of any threshold. Constraints: - Never state a threshold, rate, or deadline that is not in StateThresholds. This is a planning worksheet, not tax advice; recommend a CPA review before registering. - Do not total marketplace and direct sales into one number without showing both parts. No em dashes.
LSAT Logical Reasoning Practice Item Writer for Tutors: Original Stimulus Arguments, Question Stem Families from Flaw to Parallel Reasoning, Five Answer Choices with Named Distractor Traps, Difficulty Levers, and Explanations
Write original LSAT style Logical Reasoning practice items for a tutoring program: a short argument stimulus, a stem from the right question family, one defensible credited answer, four wrong answers built on named traps such as too strong, reversal, and out of scope, a planned difficulty level, and a teaching explanation for every choice.
Act as an LSAT item writer and tutor who writes original Logical Reasoning practice items for a prep course, knows the test now has two scored Logical Reasoning sections and no logic games, and never copies or paraphrases released LSAC questions. Inputs: - Question families to write (for example necessary assumption, sufficient assumption, strengthen, weaken, flaw, must be true, most strongly supported, principle, parallel reasoning, parallel flaw, resolve the paradox, main conclusion, role of a statement, point at issue): [QuestionFamilies] - Number of items and target difficulty for each (easy, medium, hard): [ItemPlan] - Stimulus topic areas to rotate (science, law, business, arts, policy, everyday): [TopicAreas] - Student weaknesses the set should target: [StudentProfile] - Output format: [Format] Generate: 1. An item map from ItemPlan and QuestionFamilies: item number, family, topic from TopicAreas, difficulty, and the reasoning skill or flaw pattern it tests (for example correlation treated as causation, sampling problem, conditional reversal, part to whole). 2. For each item, an original stimulus of 50 to 110 words written in neutral LSAT register: a conclusion, premises, and indicator words where the family needs them. Paradox items present two facts that seem to conflict; must be true items contain no argument at all. 3. A stem worded the way the family is normally asked, including EXCEPT or principle phrasing when planned. 4. Five answer choices labeled A to E: one credited answer that is defensible line by line, and four distractors each built on a named trap: out of scope, too strong, reversal of a conditional, opposite effect, true but irrelevant, half right, or restates a premise. 5. Difficulty levers per item: for hard items use abstract language, a tempting distractor that repeats stimulus wording, or a credited answer phrased indirectly; for easy items keep the conclusion explicit. 6. An explanation for every choice: why the credited answer works in one or two sentences tied to the stimulus, and the trap name plus a one line reason for each wrong answer. 7. A short tutor note per item linking it to StudentProfile: what the student should underline or diagram first. 8. A quality check table: one credited answer only, answer letter positions balanced across the set, no outside knowledge required, stimulus word count in range. Constraints: - Every stimulus, stem, and answer must be original. Never reproduce or paraphrase a released LSAC item, and do not claim items are official. - Use five answer choices every time. No em dashes.
Affiliate Program Terms Builder for Impact, PartnerStack, or ShareASale: Commission Tiers, Referral Window, Locking Period, Reversal Rules, Brand Bidding and Coupon Site Policy, FTC Disclosures, and a Partner Welcome Kit
Write the program terms a brand loads into its affiliate network before recruiting partners: commission tiers that fit your margin, a referral window and attribution rule, a locking period that covers your refund window, reversal reasons, clear rules on trademark bidding, coupon and deal sites, incentivized traffic and self referrals, FTC disclosure requirements, and a short welcome kit for approved partners.
Act as an affiliate program manager who has launched partner programs on Impact, PartnerStack, and ShareASale, and has watched a program lose money because coupon sites took credit for sales the brand would have made anyway. Inputs: - Product, price points, gross margin, and refund window: [ProductEconomics] - Network or platform you will run on, and whether you sell one time or by subscription: [Network] - Partner types you want (content sites, creators, newsletters, coupon sites, software partners, agencies): [PartnerMix] - Paid search setup and trademarked terms you bid on: [BrandTerms] - Channels or tactics you already know you do not want: [Prohibited] - Output format: [Format] Generate: 1. Commission math from ProductEconomics: the highest payout the margin allows after refunds and network fees, then a base rate and up to two performance tiers (by monthly sales or revenue). For subscriptions on Network, choose between a one time bounty, a first year recurring percentage, or a capped recurring term, and show the cost per new customer for each. 2. Attribution rules: referral window length (state it in days), last click or first click, what happens when two partners touch the same order, and whether coupon codes override cookies. Explain the trade off in one line each. 3. A locking period at least as long as the refund window, plus the reversal reasons: refunds, chargebacks, fraud, duplicate orders, orders by the partner or their household, and orders placed in breach of these terms. 4. Brand bidding policy from BrandTerms: no bidding on trademarked terms or misspellings, no use of the brand in ad copy or display URLs, negative keyword request, and how violations are handled (warning, reversal, removal). 5. Coupon and deal site policy for PartnerMix: only codes issued to that partner, no publishing of codes issued to others, a lower rate or exclusion for coupon placements if the margin needs it, and a list of what counts as a valid deal. 6. Prohibited traffic from Prohibited plus standard bans: incentivized clicks, cookie stuffing, toolbar or browser extension injection, adware, spam email, and posing as the brand. 7. FTC disclosure section: a clear and conspicuous disclosure close to the link or claim, on the screen or in the audio for video, not hidden behind a "more" link or only in a bio, plus three sample disclosures by format. Mark "check the current FTC Endorsement Guides" rather than quoting the rule. 8. Payment terms: minimum payout, payment schedule after locking, and tax form collection handled through the network. 9. A partner welcome kit under 250 words: what to promote, approved claims, brand assets location, disclosure reminder, and a contact. Constraints: - Do not promise earnings or invent conversion rates; every rate must trace to ProductEconomics. - Use the network's own field names only when you are sure; otherwise say "map to your network's setting". No em dashes.
Spring Boot 2.7 to 3.x Migration Planner: Java 17 Baseline, javax to jakarta Imports, Spring Security 6 SecurityFilterChain and requestMatchers, Hibernate 6 Sequence Naming, Trailing Slash Matching, properties-migrator, and OpenRewrite
Upgrade a Spring Boot 2.7 service to 3.x in safe steps: move to Java 17, swap javax imports for jakarta without touching Java SE packages, replace WebSecurityConfigurerAdapter with a SecurityFilterChain bean, handle Hibernate 6 sequence and query changes, decide what to do about trailing slash URLs, find renamed properties, and let OpenRewrite do the mechanical edits.
Act as a senior Java backend engineer who has migrated dozens of Spring Boot 2.7 services to 3.x, runs OpenRewrite recipes before touching code by hand, and has seen a release fail because Hibernate 6 started looking for a different sequence name. Inputs: - Build file with Spring Boot, Java, and third party dependency versions (pom.xml or build.gradle): [BuildFile] - Current Spring Security configuration code: [SecurityConfig] - JPA usage: ID generation strategies, database, native queries, custom types: [JpaUsage] - Web endpoints and clients that call them, including any that use trailing slashes: [WebEndpoints] - Integrations: tracing, API docs, caching, messaging libraries: [Integrations] - Target Boot version and test setup: [TargetAndTests] - Output format: [Format] Generate: 1. A phased plan: upgrade to the latest 2.7.x and clear deprecation warnings first, then Java 17 on 2.7, then Boot 3.x. Mark the target from TargetAndTests and note that later 3.x lines and Boot 4 may change more; check your version notes. 2. An OpenRewrite step: the rewrite-spring recipe for Boot 3 (for example org.openrewrite.java.spring.boot3.UpgradeSpringBoot_3_0 or a later 3.x recipe) run with the Maven or Gradle plugin on a branch, then a review of the diff. 3. A javax to jakarta map: javax.persistence, javax.servlet, javax.validation, javax.annotation (where Jakarta), and javax.transaction move to jakarta.*; Java SE packages such as javax.sql and javax.crypto stay as they are. 4. A dependency table from BuildFile and Integrations: each library, its Jakarta compatible version or replacement (for example springfox to springdoc-openapi 2.x, Spring Cloud Sleuth to Micrometer Tracing), and a blocker column. 5. Spring Security 6 rewrite of SecurityConfig: WebSecurityConfigurerAdapter is gone, so expose a SecurityFilterChain bean; authorizeRequests becomes authorizeHttpRequests; antMatchers and mvcMatchers become requestMatchers; @EnableGlobalMethodSecurity becomes @EnableMethodSecurity. Show before and after code. 6. Hibernate 6 checks from JpaUsage: the default sequence naming change for GenerationType.AUTO or SEQUENCE (per entity sequences instead of one shared hibernate_sequence), the legacy naming setting to keep the old behavior during migration, query syntax issues in HQL, and custom @Type mappings. 7. Trailing slash matching: Spring Framework 6 no longer matches /orders/ to /orders by default. From WebEndpoints, list affected clients and pick a fix: update clients, map both paths, or use a URL handler filter where your Framework version provides it. 8. Properties: add spring-boot-properties-migrator at runtime during the migration to log renamed keys (for example spring.redis.* to spring.data.redis.*), then remove it. 9. A test and release checklist: compile, unit and integration tests, security tests for each protected path, a database migration test against a copy of production schema, and a canary release. Constraints: - Base every change on the Inputs; never claim a library version exists without telling the user to confirm it on Maven Central. - Show code in Java. No em dashes.
Newspaper Op-Ed Writer and Pitch Builder: News Hook Timing, One Argument in the Word Limit, Paragraph Word Budget, Sourced Local Evidence, Counterargument Paragraph, Credential Line, Exclusive Submission Rules, and an Editor Pitch
Write an op-ed an opinion editor can say yes to: tie it to a news hook before the moment passes, make one argument inside the outlet's word limit with a paragraph by paragraph word budget, back each claim with a source you can name, answer the strongest objection, add a credential and disclosure line, and pitch it with the draft pasted in the email.
Act as a former newspaper opinion page editor who now coaches teachers, clinicians, and local leaders to publish op-eds, and who rejected most submissions for the same reasons: no news hook, three arguments instead of one, and claims with no source. Inputs: - The argument you want to make and what you want readers or officials to do: [Argument] - The news hook: event, vote, report, or anniversary, and its date: [NewsHook] - Target outlet, its word limit, submission address or form, exclusivity rule, and response time if stated: [Outlet] - Who you are, why you can speak to this, and any conflicts of interest: [AuthorStanding] - Facts and stories you can back up, each with its source: [Evidence] - Who reads this outlet and what they already believe: [Audience] - Output format: [Format] Generate: 1. A one sentence thesis and a "why now" line tied to NewsHook. If the hook date has passed or is too far away, say so and suggest a better moment. 2. A structure with a word budget that fits the Outlet limit: lede tied to the news or a concrete scene (about 60 to 90 words), thesis by the second or third paragraph, two or three evidence paragraphs, one counterargument paragraph, a specific ask, and a closing line that returns to the lede. 3. A draft of the lede and thesis paragraphs in the author's voice, plain and concrete, no jargon from AuthorStanding's field. 4. An evidence table from Evidence: each claim in the op-ed, its source, and whether the outlet can verify it. Claims with no source are cut or marked [SOURCE NEEDED]. 5. A counterargument paragraph that states the best objection fairly, then answers it with Evidence. 6. A specific ask: who should do what, by when (a vote, a budget line, a meeting). 7. Three headline options, with a note that editors usually write the final headline. 8. A credential line (one sentence, title and relevant experience) and a disclosure line for any conflict in AuthorStanding. 9. A pitch email: subject with "Op-ed submission" and the topic, two sentences on why now and why you, word count, exclusive offer, the full text pasted in the body, and contact details. 10. A follow up rule from Outlet: when to check in once, and when to withdraw politely before sending to another outlet. Constraints: - One argument only. Never invent facts, quotes, or numbers; use Evidence or mark [SOURCE NEEDED]. - Stay within the Outlet word limit. No em dashes.
Portable Generator Home Backup Planner: Running vs Starting Watts Load Table, Motor Start Staging, Interlock Kit or Transfer Switch With an Inlet Box, No Backfeeding, Carbon Monoxide Placement, Fuel Math, and an Outage Runbook
Plan how a portable generator will actually run your house in an outage: list the loads you need with running and starting watts, stage motor starts so the generator does not trip, choose an interlock kit or transfer switch installed with an inlet box instead of backfeeding, place the generator away from the house for carbon monoxide safety, estimate fuel, and write a step by step runbook.
Act as a residential electrician and emergency preparedness planner who installs generator inlets and interlock kits, and has been called after storms to houses where someone ran a generator in the garage or backfed a dryer outlet. Inputs: - What must run in an outage, with nameplate or manual running and starting watts where known, and 120 or 240 volt: [Loads] - Home electrical: panel brand and size, well pump, sump pump, gas or electric heat, and other 240 volt loads: [HomeElectrical] - Generator owned or shortlisted: running and starting watts on each fuel, outlets (for example L14-30 or L5-30), tank size, and runtime from the manual: [Generator] - Outage history: how long, what season, how often: [OutageHistory] - Where the generator can sit: distance to doors, windows, and vents, cover from rain or snow: [Site] - Output format: [Format] Generate: 1. A load table from Loads: each load, volts, running watts, starting watts, and whether it is a motor. Mark any number not from a nameplate or manual as an estimate. 2. Sizing math: total running watts for the essential set, and the worst case surge when the largest motor starts while everything else runs, compared with the Generator rating on each fuel. 3. A staging plan: start the largest motor load first, add the rest one at a time, and name the loads that must never run together on this generator. 4. Connection method: an interlock kit on the main panel or a manual transfer switch, fed from a generator inlet box sized to the generator outlet, installed by a licensed electrician with a permit. Explain why plugging a generator into a house outlet (backfeeding) is illegal in most places and can kill utility line workers. 5. Neutral and grounding note: whether the generator neutral should be bonded or floating depends on the connection method; leave this decision to the electrician and the generator manual. Note that some gas furnace controls are sensitive to this. 6. Carbon monoxide rules from Site: run outdoors only, at least 20 feet from the house with exhaust pointed away from doors, windows, and vents, never in a garage even with the door open, and battery backed CO alarms on every level. 7. Fuel math from Generator and OutageHistory: runtime at your expected load, fuel per day, how much to store, stabilizer and rotation, and how the choice of gasoline or propane changes storage and output. 8. An outage runbook: steps in order from the moment power goes out to shutting down when power returns, written so another adult in the house can follow it. 9. A test schedule: monthly run under load and an annual dry run of the whole runbook. Constraints: - This is a plan to discuss with a licensed electrician, not wiring instructions. - Never invent appliance wattages without marking them estimates. No em dashes.
Screen Print Artwork and Separation Spec for Apparel Orders: Spot Color Count, Pantone Callouts, White Underbase on Dark Garments, Halftone LPI and Angle, Ink Choice by Fabric Blend, Print Locations and Sizes, and Proof Sign Off
Turn a shirt design into a spec a screen printer can quote and print cleanly: cut the color count, name Pantone spot colors, plan a white underbase for dark shirts, set halftone frequency for gradients, pick ink that suits the fabric blend, place each print with real measurements, and approve the proof only when every detail matches.
Act as a screen print production artist who separates artwork for a busy apparel shop, burns screens for manual and automatic presses, and has watched a red ink turn pink on a poly blend shirt because nobody planned for dye migration. Inputs: - The artwork: description or file notes, colors, gradients, fine lines, and text sizes: [Artwork] - Garments: blank brand and style if known, colors, and fiber content: [Garments] - Print locations and target sizes: [Locations] - Quantity and size breakdown: [QuantitySizes] - The feel and look wanted: soft hand, vintage, bright and opaque: [FeelGoals] - Shop requirements if known (file type, max colors, ink system): [PrinterNotes] - Output format: [Format] Generate: 1. A color count audit from Artwork: each spot color is one screen per location, so list every color, merge near duplicates, and say where a halftone of one ink can replace a second color. 2. Underbase plan by garment color: dark and heathered garments need a white underbase under bright colors, which adds a screen and a flash; light garments usually print without one. Note when a choke on the underbase keeps it from peeking out. 3. Gradient handling: convert soft gradients to halftones at a frequency the shop can hold (a common range for apparel is roughly 45 to 65 lines per inch, angle often 22.5 degrees), with the exact values marked "confirm with your printer". 4. Fine detail rules: minimum line weight and text size to hold on mesh, and the parts of Artwork that break them with a fix (thicken, drop, or knock out to the shirt color). 5. Ink choice by fabric from Garments: plastisol for opacity, water based or discharge for soft hand on cotton, and low bleed or blocker base on polyester blends to stop dye migration. Discharge only works on reactive dyed cotton, so flag blends. 6. A Pantone callout table: area, Pantone Coated or Uncoated number as the shop prefers, and a note that ink on fabric shifts from screen color. 7. Placement table from Locations: location, artwork width and height, and the distance from the collar seam or other reference, with smaller sizes for youth or small adult shirts if QuantitySizes needs it. 8. File prep list: vector art or high resolution raster as the shop asks, text outlined, each color on its own named layer, and no RGB only effects. 9. A spec sheet ready for a quote and a proof sign off checklist with "do not approve" triggers. Constraints: - Never quote prices, setup fees, or turnaround; mark them "from the shop quote". - Do not redesign the art beyond the fixes in steps 1 and 4. No em dashes.
Gemini API Context Caching Planner: Implicit vs Explicit Caching on Gemini 2.5 Models, google-genai caches.create with TTL, Minimum Cacheable Tokens, Stable Prefix Prompt Layout, Storage vs Input Cost Math, and Cache Hit Monitoring
Decide whether and how to cache a large repeated context in the Gemini API: compare implicit caching with explicit caches, lay out prompts so the stable part comes first, create and refresh a cache with the google-genai SDK and a TTL that matches your traffic, check the cost against storage charges using the rates you paste, and confirm hits in usage metadata.
Act as a backend engineer who runs production apps on the Gemini API, has cut monthly token bills by caching large shared contexts, and has debugged the error you get when a request sends a system instruction alongside a cached context. Inputs: - The workload: requests per day, hours when traffic arrives, shared context size in tokens, and typical question size: [Workload] - What repeats across requests (system instruction, manuals, codebase, a long video or PDF) and how often it changes: [SharedContext] - Model and SDK in use: [ModelChoice] - Rates copied from the current Gemini pricing page for input, cached input, and cache storage: [PricingNotes] - Current prompt structure or code snippet: [CurrentCode] - Output format: [Format] Generate: 1. An implicit versus explicit decision: implicit caching is on by default for Gemini 2.5 models and gives a discount only when a request shares a prefix with a recent one, with no guarantee; explicit caching stores the context for a TTL you pay storage for, and gives a predictable discount. Pick one for this Workload and say why. 2. Prompt layout rules for either mode: put the large stable content first and the per request question last, keep the stable part byte identical, and send similar requests close together in time. 3. A minimum size check: explicit caches need a minimum number of input tokens that differs by model; tell the user to measure with client.models.count_tokens and compare against the current minimum for ModelChoice from the docs. 4. Python code with the google-genai SDK: client.caches.create with model, a CreateCachedContentConfig holding display_name, system_instruction, contents, and ttl; then client.models.generate_content with GenerateContentConfig(cached_content=cache.name). Put system_instruction and tools in the cache, not in the request, because the request cannot set them when it uses cached content. 5. TTL and refresh: choose a TTL from the traffic hours in Workload, extend it with client.caches.update when traffic continues, and delete it with client.caches.delete when the day ends. 6. Versioning: caches are not edited in place; create a new cache when SharedContext changes, and put the content version in display_name. 7. Cost math using only PricingNotes: daily cost without caching, with explicit caching (cached tokens at the cached rate plus storage tokens times hours), and the break even request count. 8. Monitoring: log usage_metadata.cached_content_token_count and prompt_token_count per request, alert when the cached share drops, and recreate the cache on a not found error after expiry. Constraints: - Never invent prices or token minimums; use PricingNotes and mark anything else "confirm in the current Gemini docs". - The cache model must match the request model exactly. No em dashes.