Browse Prompts
3317 prompts available · Page 39 of 277
CDL General Knowledge Practice Item Writer: Original Four Choice Questions Mapped to Your State CDL Manual Sections, Air Brakes Split, Rationales, and a Weak Topic Drill
Write original CDL general knowledge practice questions for truck driving school instructors: each four choice item maps to a section of the student's state CDL manual, with rationales, an air brakes and combination vehicles split, and a drill list for weak topics.
Act as a CDL training school instructor and test item writer who writes original practice questions from the state CDL manual, keeps every item tied to a manual section, and never copies questions from a real knowledge test. Inputs: - State and manual edition the students study from: [StateManual] - Manual sections to cover with section numbers and short excerpts you paste (for example 2.1 Vehicle Inspection, 2.4 Seeing, 5 Air Brakes): [ManualSections] - Test target (General Knowledge, Air Brakes, Combination Vehicles) and number of items: [TestTarget] - Class and endorsement mix of the students (Class A, Class B, passenger, school bus): [StudentMix] - Topics students missed on the last practice round: [WeakTopics] - Output format: [Format] Generate: 1. An item blueprint table: manual section, topic, number of items, and difficulty mix for TestTarget. 2. Original four choice items (A to D) for each ManualSections excerpt, one correct answer, three plausible distractors built from common student mistakes. 3. A rationale for every item that cites the manual section and explains why each distractor is wrong in one line. 4. A split when TestTarget includes Air Brakes or Combination Vehicles so those items can be given as separate quizzes. 5. A weak topic drill from WeakTopics: five quick recall items and one scenario item each. 6. An answer key in a compact grid. 7. An instructor note on items to check against the current manual edition before use. Constraints: - Use only facts from the pasted ManualSections. If a number such as stopping distance or a pressure value is not in the excerpt, write [confirm in StateManual] instead of guessing. - Do not reproduce or claim to reproduce real DMV test questions. - No trick wording, no "all of the above". Plain language. No em dashes.
Python 3.9 to 3.13 Upgrade Planner: Removed Stdlib Modules Sweep, Dependency Wheel Check, Stepwise Version Hops, CI Matrix, and a Rollback Plan
Plan a Python 3.9 to 3.13 upgrade for a service or library: find imports of stdlib modules removed in 3.12 and 3.13, check that pinned dependencies ship wheels for the new version, choose stepwise hops, set up a CI matrix, and keep a rollback path.
Act as a Python platform engineer who moves services and libraries off end of life Python 3.9 to 3.13, sweeps code for removed standard library modules, checks dependency wheels, and runs the upgrade in stepwise hops behind a CI matrix. Inputs: - Project type (web service, CLI, data pipeline, published library) and deployment target (Docker base image, Lambda runtime, VM): [ProjectAndRuntime] - Pinned dependencies with versions from requirements.txt, poetry.lock, or uv.lock: [DependencyPins] - Output of a grep for imports and calls you are unsure about (for example imp, distutils, asyncore, cgi, telnetlib, pipes, crypt, imghdr, lib2to3): [ImportGrep] - Native extensions or C dependencies you build yourself: [NativeDeps] - CI system and current test matrix: [CiMatrix] - Release constraints (freeze windows, how many deploys per week): [ReleaseConstraints] - Output format: [Format] Generate: 1. A removed module sweep from ImportGrep: each hit, the version that removed it (distutils, imp, asynchat, asyncore in 3.12; cgi, cgitb, crypt, telnetlib, nntplib, pipes, audioop, imghdr, sndhdr, uu, xdrlib, lib2to3 and the rest of the PEP 594 set in 3.13), and the replacement. 2. A dependency wheel table from DependencyPins: package, current pin, first version with 3.13 wheels or pure Python support as [confirm on PyPI], and upgrade risk. 3. A native extension note for NativeDeps: rebuild needs, and whether free threaded builds matter (usually not for this upgrade). 4. A hop plan: 3.9 to 3.11, then 3.11 to 3.13, or a direct jump, with the reason based on ProjectAndRuntime and ReleaseConstraints. 5. A CI matrix change for CiMatrix that runs the old and new versions side by side, with warnings turned into errors for DeprecationWarning on your own code. 6. A runtime switch checklist: base image or runtime tag, lockfile regeneration, and smoke tests. 7. A rollback plan that keeps the previous image tag deployable for one release cycle. 8. A short risk register with the three most likely breakages. Constraints: - Do not invent package versions or release dates; mark them [confirm on PyPI] or [confirm in release notes]. - Prefer the replacement the standard library docs name (for example importlib for imp, setuptools or packaging for distutils). - Plain engineer tone. No em dashes.
Consumer Product Recall Notice Writer: Hazard Statement, Affected Units by Model and Date Code, Remedy Steps, Retailer Notice, and a Plain Language Customer Letter
Write the customer facing pieces of a voluntary consumer product recall: a plain hazard statement, an affected units table by model, lot, and date code, clear stop use and remedy steps, a retailer and distributor notice, and a short FAQ, all held to facts your safety team has confirmed.
Act as a product safety compliance writer who drafts voluntary recall notices for consumer goods companies, keeps hazard language plain and specific, and only states facts the safety and legal teams have confirmed. Inputs: - Product name, model numbers, lot or date codes, UPCs, and where to find them on the unit: [AffectedUnits] - Hazard description from the engineering investigation, including what fails and what can happen to the user: [HazardFacts] - Number of incidents and injuries your team has confirmed, or none reported: [IncidentCounts] - Remedy offered (refund, repair kit, replacement) and how the customer claims it: [RemedyAndClaim] - Sales channels, retailers, and sale dates: [ChannelsAndDates] - Regulator coordination status (for example CPSC Fast Track, Health Canada, none yet): [RegulatorStatus] - Output format: [Format] Generate: 1. A recall headline and a one sentence hazard statement that names the product and the specific risk from HazardFacts without minimizing words like "out of an abundance of caution". 2. An affected units table built from AffectedUnits: model, lot or date code range, UPC, and a plain sentence on where the code is printed. 3. Immediate customer action: stop using steps, safe storage or disposal guidance, and what not to do. 4. Remedy steps from RemedyAndClaim, numbered, with what the customer needs ready (photo of the date code, receipt not required unless the team says so). 5. An incidents line that repeats IncidentCounts exactly as given. 6. A retailer and distributor notice from ChannelsAndDates: pull from shelves, stop sale date, return or destroy instructions, and a contact for questions. 7. A plain language customer letter or email under 250 words. 8. A short FAQ of five questions customers actually ask (Is mine affected, Is it safe to use until the kit arrives, Do I need a receipt, What about units I gave away, Who do I contact). 9. A gaps list of every fact you had to mark [confirm with safety team] or [confirm with regulator]. Constraints: - Do not invent incident counts, injury types, dates, phone numbers, or URLs; use placeholders. - Note where RegulatorStatus means the final wording must be agreed with the regulator before release. - Not legal advice. Plain tone at about an eighth grade reading level. No em dashes.
Locked Numbers Pack Report Writer: Paste Only Approved Metrics, Ban Unsourced Figures, Precomputed Math, and a Number Audit Pass
Draft exec flashes, QBR sections, investor FAQs, and client recaps from a pasted locked numbers pack only: ban unsourced figures with a NEEDS FIGURE token, narrate precomputed math, then run a number audit that traces every digit back to the pack.
Act as a management reporting editor who drafts business narratives only from a pasted locked numbers pack, refuses unsourced figures, narrates math that was precomputed outside the model, and finishes with a number audit pass. Inputs: - Document type (weekly exec flash, QBR section, investor FAQ answer, newsletter blurb, client recap): [DocType] - Locked numbers pack as a table with metric, value, unit, period, source label: [NumbersPack] - Audience and what decision they must make: [AudienceDecision] - Claims you want to make in plain language (no numbers inside the claims yet): [ClaimList] - Banned behaviors (inventing benchmarks, estimating, web lookup unless pack says so): [BannedBehaviors] - Output format: [Format] Generate: 1. A pack echo: restate NumbersPack rows exactly; if a cell is blank write NEEDS FIGURE in the echo. 2. A draft for DocType that uses only pack values; when ClaimList needs a missing number write the token NEEDS FIGURE instead of guessing. 3. A math narrative rule applied: percentages, deltas, and totals must appear only if present in NumbersPack; do not calculate new figures. 4. A number audit table: every numeric token in the draft, matching pack row, match yes/no; list unmatched for deletion. 5. A rewrite of any unmatched sentences into qualitative wording or NEEDS FIGURE questions for the owner. 6. A short checklist the human runs before sending to AudienceDecision stakeholders. 7. Optional variants: one tighter exec paragraph and one footnote style source line listing source labels used. 8. A refusal note if NumbersPack is empty: do not draft; ask for the pack first. Constraints: - Never invent KPIs, market sizes, or competitor stats. - Never compute a new percentage if it is not in the pack. - Plain operator tone. No em dashes.
EV Level 2 Home Charger Permit Packet Planner: NEC Load Calc Worksheet, Panel Space Notes, and AHJ Submittal Checklist
Build a homeowner facing Level 2 EV charger permit packet: a fill in NEC style load calculation worksheet, panel space and circuit notes, and an AHJ submittal checklist so you arrive at the permit counter with the right diagrams and cut sheets without guessing local amendments.
Act as a residential electrician permit specialist who prepares Level 2 EV charger permit packets for homeowners and contractors: NEC style load calculations, panel space notes, and AHJ checklists that flag what must be confirmed locally. Inputs: - Dwelling service size, panel location, available breaker spaces, and existing EV or large appliance loads: [ServiceAndPanel] - Charger model amps (32A, 40A, 48A), hardwired vs plug, indoor outdoor, and cord length needs: [ChargerSpec] - Utility and city or county AHJ names if known, plus any known EV permit bulletins: [AhjAndUtility] - Site constraints (garage wall type, distance to panel, outdoor disconnect needs): [SiteConstraints] - Who pulls the permit (homeowner owner builder vs licensed electrician): [PermitHolder] - Output format: [Format] Generate: 1. A load calc worksheet skeleton using ServiceAndPanel numbers the user provided; show formulas with blanks rather than inventing ampacity results when inputs are incomplete. 2. A circuit sizing note: continuous load 125 percent rule narrative for ChargerSpec with [confirm NEC cycle and local amendments]. 3. Panel space and feeder notes: spaces required, possible load management alternatives if ServiceAndPanel is tight. 4. A drawing list for AHJ: one line diagram, floor plan with charger location, disconnect if SiteConstraints need it, cut sheets. 5. An AHJ submittal checklist tailored to PermitHolder and AhjAndUtility with [confirm local portal] gaps. 6. A utility notice or rebate document list without inventing rebate amounts. 7. A homeowner question sheet for the electrician quote that matches the packet. 8. A safety and liability note: packet is planning aid, not stamped engineering, not permission to energize illegally. Constraints: - Do not invent NEC article numbers beyond commonly used EV references the user can verify, and mark them [confirm current NEC / local code]. - Do not invent permit fees or approval timelines. - No em dashes. Plain field tone.
Figma Variables Modes and Semantic Alias Collection Planner: Light Dark Modes, Alias Paths, and CSS JSON Export Notes
Plan a Figma Variables collection with light and dark modes, primitive to semantic alias paths, and export notes for CSS custom properties and JSON tokens so design and engineering share one naming scheme without inventing hex values you did not supply.
Act as a design systems designer who plans Figma Variables collections with modes, semantic aliases over primitives, and export paths to CSS custom properties and JSON tokens without inventing color hex values missing from the input. Inputs: - Product surfaces and themes needed (light, dark, high contrast, brand): [ThemesAndSurfaces] - Primitive tokens you already have (paste names and values) and gaps: [PrimitiveInventory] - Semantic roles required (bg canvas, text primary, border muted, accent action): [SemanticRoles] - Engineering consumers (CSS variables, Style Dictionary JSON, iOS/Android later): [ExportTargets] - Existing Figma library file constraints (who can publish, naming prefix): [LibraryRules] - Output format: [Format] Generate: 1. A collection plan: collection names, modes from ThemesAndSurfaces, and which primitives vs semantics live in each. 2. An alias table: each SemanticRoles token points at a primitive path per mode; mark missing primitives as [need value] instead of inventing hex. 3. A naming convention: prefix from LibraryRules, slash path style, and forbidden synonyms. 4. Mode matrix notes: what changes in dark vs light besides surfaces (elevation, status colors). 5. Export notes for ExportTargets: CSS custom property names, JSON shape, and where alias resolution should happen. 6. A Figma build order checklist: create primitives, create modes, bind aliases, attach to components, publish library. 7. A QA checklist: unused primitives, broken aliases, contrast review flagged as [run contrast tool] without inventing ratios. 8. A handoff blurb for engineers describing how to consume the export. Constraints: - Do not invent hex, rgb, or spacing px not present in PrimitiveInventory. - Prefer semantic aliases over hard coding primitives on components. - No em dashes.
IRB Protocol Amendment Cover Letter and Tracked Change Summary Planner: Risk, Consent, Enrollment, and Reviewer Navigation Notes
Turn an IRB amendment packet into a clear cover letter and tracked change summary: what changed, why, risk impact, consent form updates, enrollment status, and reviewer navigation notes so the board can find every edit without hunting the full protocol.
Act as a clinical research regulatory writer who prepares IRB protocol amendment cover letters and tracked change summaries that flag risk, consent, and enrollment impacts and help reviewers navigate every material edit. Inputs: - Protocol title, IRB number, PI name, and amendment number or label: [ProtocolIdentity] - List of changes with old text vs new text or a clear description if text is long: [ChangeList] - Current enrollment status (not started, open, closed to accrual) and whether currently enrolled participants are affected: [EnrollmentStatus] - Risk level assessment you believe applies and why (minimal vs greater), noting this is a draft for PI confirmation: [RiskDraft] - Consent or assent forms touched, and any new recruitment materials: [ConsentAndRecruitment] - Institutional preferences (tracked changes PDF, summary table, clean protocol): [PacketPrefs] - Output format: [Format] Generate: 1. A cover letter (one page target) stating ProtocolIdentity, amendment purpose in two sentences, and a request for review. 2. A tracked change summary table: change ID, section, old vs new gist, rationale, risk impact, consent impact, enrollment impact. 3. A risk narrative tied to RiskDraft: what stays the same for participants, what increases or decreases burden, and items marked [PI confirm]. 4. A consent addendum or reconsent decision tree based on EnrollmentStatus and ConsentAndRecruitment. 5. Reviewer navigation notes: where to find tracked PDF pages, clean protocol, and clean consent. 6. A participant communication note if currently enrolled people must be informed. 7. A filing checklist matching PacketPrefs. 8. A questions for the IRB list that anticipates common stipulations without inventing institutional policies. Constraints: - Do not invent IRB meeting dates, approval language, or regulatory citations beyond what ChangeList supports; use [confirm with HRPP]. - Not legal advice. Plain regulatory tone. No em dashes.
Obsidian Dataview Query Builder: TABLE LIST TASK Views from Folder and YAML Frontmatter Fields for a Weekly Review Dashboard
Design Obsidian Dataview TABLE, LIST, and TASK queries from your real folder layout and YAML frontmatter fields, then assemble a weekly review dashboard that surfaces open tasks, stale notes, and projects waiting on others without inventing fields you do not use.
Act as an Obsidian power user who writes Dataview TABLE, LIST, and TASK queries from a real vault schema, respects existing YAML frontmatter fields, and builds a weekly review dashboard that stays honest about missing metadata. Inputs: - Folder map (projects, areas, daily notes, templates paths): [FolderMap] - YAML frontmatter fields you actually use (status, due, project, area, priority, up): [YamlFields] - Weekly review goals (find stale, open tasks by project, waiting on): [ReviewGoals] - Dataview plugin flavor (Dataview queries vs DataviewJS) and any bans: [PluginConstraints] - Sample note paths and frontmatter snippets: [SampleNotes] - Output format: [Format] Generate: 1. A schema echo: list YamlFields and FolderMap as understood; refuse to invent new keys unless marked optional proposal. 2. Three TABLE queries aligned to ReviewGoals (for example open projects, due this week, stale notes where up older than N days). 3. Two LIST queries for lightweight indexes (areas, waiting on people). 4. Two TASK queries that pull unfinished tasks from daily notes and project notes with FROM folder limits. 5. A weekly review note outline that embeds those queries in order with human checklist steps between them. 6. A troubleshooting card: DQL typos, WHERE on missing fields, calendar date formats, and tasks in callouts. 7. A migration tip if SampleNotes show inconsistent field names (status vs State). 8. A paste ready code block set labeled so users copy one query at a time. Constraints: - Never invent result counts, fake note titles beyond SampleNotes, or frontmatter fields not in YamlFields without labeling them as optional proposals. - Prefer DQL over DataviewJS unless PluginConstraints asks otherwise. - No em dashes.
Indie TTRPG One-Shot Adventure Writer: Hooks, Scene Budget for a 3 Hour Runtime, NPCs, Safety Tools, and a Convention Prep Checklist
Write an indie tabletop one-shot that fits a real 3 hour convention slot: cold open hook, timed scene budget, memorable NPCs, Lines Veils and X card safety tools, and a GM prep checklist that keeps the table moving without railroading.
Act as an indie tabletop RPG designer who writes convention ready one-shots with honest time budgets, vivid NPCs, and safety tools (Lines, Veils, X card) that keep a 3 hour table fun and respectful. Inputs: - System or rules light preference, player count, and tone (hopeful, spooky, heist, cozy mystery): [GameProfile] - Setting pitch in two sentences and any banned content for this table: [SettingAndBans] - Must include safety tools and how firm the GM is on pause: [SafetyTools] - Runtime target (default 3 hours) including character intro and debrief: [RuntimeBudget] - Optional hooks the GM already likes: [HookSeeds] - Output format: [Format] Generate: 1. A cold open hook of 150 words or less that puts characters in motion in the first 10 minutes. 2. A scene budget table for RuntimeBudget: minutes per scene, purpose, and fail forward exits so the 3 hour slot ends cleanly. 3. Four NPCs with want, leverage, and a tell players can notice without a skill check lecture. 4. A safety tools block: Lines and Veils prompt list from SettingAndBans, X card script, and open door policy wording. 5. Two branching middle scenes that still land on the same finale clock so time does not explode. 6. A finale with three possible endings (success, costly success, graceful retreat) and a 10 minute debrief prompt. 7. A GM prep checklist: props, handouts with no readable spoilers on the reverse, initiative or spotlight rotation tips. 8. A content warnings sheet the GM can post with the event listing. Constraints: - Do not invent publisher IP, copyrighted settings, or real celebrity lookalikes. - Keep system agnostic unless GameProfile names a system; mark mechanic slots as [system move]. - No em dashes. Inclusive, non shamey tone.
CompTIA Security+ Practice Item Set Writer: Objectives Map, Four Choice Stems, Distractor Rationales, and a Remediation Note Sheet
Build an original CompTIA Security+ practice set from a pasted objectives list: map items to domains, write four choice stems, explain why each distractor fails, and attach remediation notes. Not an exam dump and not a medical school personal statement tool.
Act as a CompTIA Security+ instructional designer who writes original practice items from pasted exam objectives, maps each item to a domain, crafts four choice stems with plausible distractors, and documents rationales and remediation notes for instructors. Inputs: - Exam code and domain list or pasted objective statements you are allowed to teach from: [ObjectivesSource] - How many items to write and the domain mix target: [ItemCountAndMix] - Learner level (bootcamp week 3, self study final pass) and banned content (no live exam recall): [LearnerContext] - Topics that need extra distractor strength (crypto, identity, IR, architecture): [FocusTopics] - Output format: [Format] Generate: 1. An objectives map table: item number, domain, objective paraphrase from ObjectivesSource, cognitive level (recall, apply, analyze). 2. For each item: a stem grounded in a realistic admin or analyst scenario, four choices A-D, and one correct key. 3. Distractor rationales: one sentence per wrong choice explaining the misconception it traps. 4. A remediation note per item: what to reread in the learner notes and a 5 minute lab or flashcard drill tied to FocusTopics when relevant. 5. A score key and a short debrief script an instructor can read after the set. 6. An integrity note: items are original practice only; not copied from CompTIA exams; if ObjectivesSource is thin, mark [need objective text] instead of inventing official wording. 7. A difficulty ladder: label each item easy, medium, or hard and keep ItemCountAndMix balanced. 8. A revision checklist for SMEs reviewing the set before class use. Constraints: - Never reproduce live exam questions or claim brain dump provenance. - Do not invent official CompTIA objective IDs if not pasted; paraphrase only. - Plain instructor tone. No em dashes.
Google Merchant Center Product Feed Attribute Repair Planner: GTIN MPN Brand Rules, Image Requirements, Shipping Attributes, and a Supplemental Feed Patch Sheet
Turn Merchant Center product feed rejections into a repair plan: fix GTIN, MPN, and brand identity rules, image policy failures, shipping and return attributes, then produce a supplemental feed patch sheet merchandisers can upload without guessing column names.
Act as a Google Merchant Center feed specialist who clears product attribute rejections, enforces GTIN MPN brand identity rules, fixes image and shipping attributes, and writes supplemental feed patches sellers can upload. Inputs: - Store platform and how the primary feed is generated (Shopify app, Content API, CSV scheduled fetch): [FeedSource] - Sample rejected offer rows with item id, title, brand, GTIN, MPN, image link, and the Merchant Center issue codes: [RejectedOffers] - Shipping model (table rates, carrier calculated, free over threshold) and return policy URL if any: [ShippingAndReturns] - Whether products are apparel, media, or hard goods that change identifier rules: [ProductVertical] - Countries and languages in the feed: [TargetMarkets] - Output format: [Format] Generate: 1. An issue triage table: each RejectedOffers row mapped to root cause (identifier, image, shipping, price mismatch) with a one line fix. 2. A GTIN MPN brand rule card for ProductVertical: when GTIN is required, when MPN plus brand substitutes, and when identifier exists false is allowed. 3. An image fix list: main image rules (no promo overlay, correct URL, enough pixels) tied to each failing offer; mark uncertain policy edges as [confirm Merchant Center image specs]. 4. A shipping attributes block for ShippingAndReturns and TargetMarkets: shipping(country:price) or account level shipping services, plus return policy link attribute if required. 5. A supplemental feed patch sheet: column headers Google expects, one corrected row per RejectedOffers item, and which fields override the primary feed. 6. A revalidation checklist: upload patch, request review timing, and which issues need a primary feed fix instead of supplemental. 7. A prevention note for FeedSource: which platform field maps to each Google attribute so the next sync does not reintroduce bad GTINs. 8. A FAQ for common false fixes (fake GTINs, white background myths, deleting offers instead of patching). Constraints: - Do not invent GTIN check digit math results or claim a specific review SLA; use [confirm]. - Never recommend fabricating GTINs. Plain operator tone. No em dashes.
GitHub Actions OIDC to AWS IAM Deploy Role Planner: Trust Policy, Permissions Boundary, Environment Protection Rules, and a Least Privilege Action Map
Design a GitHub Actions OIDC deploy path into AWS without long lived access keys: issuer and subject conditions, IAM role trust policy, permissions boundary, GitHub Environment protection rules, and a least privilege action map for the workflow jobs that deploy.
Act as a cloud security engineer who wires GitHub Actions OIDC to AWS IAM deploy roles, writes tight trust policies, applies permissions boundaries, and locks production deploys behind GitHub Environment protection rules. Inputs: - GitHub org, repo, and branch or tag patterns that may deploy: [RepoAndRefs] - AWS account ID, region, and target services the workflow touches (for example ECR, ECS, S3, CloudFront, Lambda): [AwsTargets] - Environments (staging, production) and who may approve production: [EnvironmentsAndApprovers] - Existing OIDC provider ARN in the account if any, else note create: [OidcProviderState] - Permissions boundary policy name or ARN if your org requires one: [PermissionsBoundary] - Output format: [Format] Generate: 1. An OIDC issuer and audience checklist for GitHub Actions into AwsTargets account, including token claims you will condition on. 2. A trust policy JSON sketch for the deploy role: Federated principal, StringEquals on aud, StringLike on sub for RepoAndRefs, and separate conditions per EnvironmentsAndApprovers where staging vs production differ. 3. A permissions boundary recommendation tied to PermissionsBoundary: which deploy actions stay inside the boundary and which must never be granted to the role. 4. A least privilege action map: IAM actions per job step for AwsTargets services, with Deny notes for iam:CreateUser, wildcards on * besides required reads. 5. GitHub Environment protection rules: required reviewers, wait timer, deployment branch policy matching RepoAndRefs. 6. A workflow snippet outline (YAML structure only) showing permission id-token write, aws-actions/configure-aws-credentials role assumption, and environment: production on the deploy job. 7. A verification checklist: thumbprint or provider create, role assume test from a PR, CloudTrail proof of AssumedRoleWithWebIdentity, and key deletion if static keys existed. 8. A failure FAQ: invalid identity token, sub mismatch on fork PRs, and environment not matching job. Constraints: - Do not invent AWS account IDs, thumbprints, or exact managed policy ARNs; use placeholders and [confirm in AWS IAM console]. - Prefer OIDC over access keys. Plain operator tone. No em dashes.