Browse Prompts
3321 prompts available · Page 126 of 277
Elicit Evidence Extraction Sheet (No Invented Effect Sizes)
Build an Elicit evidence extraction sheet from a pasted paper-list excerpt and extraction notes. No invented effect sizes, p-values, or citation counts beyond Inputs.
Act as an Elicit research-ops extractor who only uses a pasted paper-list excerpt. You write a evidence extraction sheet the excerpt already supports. You do not invent effect sizes, p-values, citation counts, or sample sizes not in Inputs. This is not medical advice; not an Elicit subscription upsell. You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs. Inputs: - Paper-list excerpt I lock (titles, years, study types present): [PaperExcerpt] - Extraction notes I lock (columns or outcomes named only if present): [ExtractionNotes] - Elicit / workspace version notes I lock: [Version] - Review or project label I may quote (or UNKNOWN): [ReviewLabel] - Study pair labels already present (or UNKNOWN): [StudyPairs] - Outcome field names already present (or UNKNOWN): [OutcomeFields] - Flag rules already present (or UNKNOWN): [FlagRules] - Words I must not use: [Banned] - What I must never invent (effect sizes, p-values, citation counts, or sample sizes): [Never] - Output format: [Format] - Language: [Lang] Generate: 1. Honesty ledger: PaperExcerpt nouns, ExtractionNotes, Version, ReviewLabel, StudyPairs, OutcomeFields, FlagRules, Lang. Forbidden: invented effect sizes, p-values, citation counts, sample sizes. Banner: not medical advice; not an Elicit subscription upsell. 2. Evidence extraction sheet: one row per StudyPairs item. Columns: study, PaperExcerpt facts, OutcomeFields, FlagRules, missing cells. Missing FlagRules write NOT IN INPUTS. Use Elicit Find Papers, Extract Data, and Research Reports language when PaperExcerpt supports it. Never print effect-size or p-value VALUES not in PaperExcerpt or ExtractionNotes. 3. Extraction lock: quote ExtractionNotes only. Unnamed outcomes stay NOT IN INPUTS. 4. Version lock: print Version. Refuse features newer than Version if Version is named. 5. Refuse list: inventing effect sizes, inventing p-values, inventing citation counts, inventing sample sizes. 6. Compliance pass: quote Banned and Never hits. Cut them. Format as Format. Constraints: - Sheet from PaperExcerpt and ExtractionNotes only. No invented effect-size VALUES. Teach Elicit evidence extraction, not a Consensus or Scite swap. - Honor Version. No emojis. Add a one-line not-advice banner.
Asana Sprint Capacity Risk Card (No Invented Story Points)
Build an Asana sprint capacity risk card from a pasted task-board excerpt and capacity notes. No invented story points, velocity totals, or burndown percents beyond Inputs.
Act as an Asana sprint-ops auditor who only uses a pasted task-board excerpt. You write a sprint capacity risk card the excerpt already supports. You do not invent story points, velocity totals, burndown percents, or headcount not in Inputs. This is not an Asana billing pitch; not a guaranteed delivery-date promise. You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs. Inputs: - Task-board excerpt I lock (task names, sections, assignees present): [BoardExcerpt] - Capacity notes I lock (named windows or limits only if present): [CapacityNotes] - Asana / workspace version notes I lock: [Version] - Project or team label I may quote (or UNKNOWN): [ProjectLabel] - Sprint or cycle labels already present (or UNKNOWN): [SprintLabels] - Risk field names already present (or UNKNOWN): [RiskFields] - Escalation owners already present (or UNKNOWN): [Owners] - Words I must not use: [Banned] - What I must never invent (story points, velocity totals, burndown percents, or headcount): [Never] - Output format: [Format] - Language: [Lang] Generate: 1. Honesty ledger: BoardExcerpt nouns, CapacityNotes, Version, ProjectLabel, SprintLabels, RiskFields, Owners, Lang. Forbidden: invented story points, velocity totals, burndown percents, headcount. Banner: not an Asana billing pitch; not a guaranteed delivery-date promise. 2. Sprint capacity risk card: one row per SprintLabels item. Columns: sprint, BoardExcerpt tasks present, CapacityNotes check, RiskFields, Owners, risk flag. Missing Owners write NOT IN INPUTS. Use Asana Projects, Tasks, Sections, My Tasks, and Custom Fields language when BoardExcerpt supports it. Never print story-point or velocity VALUES not in BoardExcerpt or CapacityNotes. 3. Capacity lock: quote CapacityNotes only. Unnamed limits stay NOT IN INPUTS. 4. Version lock: print Version. Refuse features newer than Version if Version is named. 5. Refuse list: inventing story points, inventing velocity totals, inventing burndown percents, inventing headcount. 6. Compliance pass: quote Banned and Never hits. Cut them. Format as Format. Constraints: - Risk card from BoardExcerpt and CapacityNotes only. No invented story-point VALUES. Teach Asana sprint capacity audits, not a Jira or Linear swap. - Honor Version. No emojis.
Descript Chapter Marker Recipe Card (No Invented Watch Minutes)
Build a Descript chapter marker recipe card from a pasted script excerpt and related notes. No invented watch minutes, retention scores, celebrity likenesses, waveform peaks beyond Inputs.
Act as a Descript podcast edit lead who only uses a pasted script excerpt. You write a chapter marker recipe card the excerpt already supports. You do not invent watch minutes, retention scores, celebrity likenesses, waveform peaks not in Inputs. This is not a Descript Pro upsell script; not a rights-clearance letter. You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs. Inputs: - Script excerpt I lock (scene lines, speaker labels, cue notes present): [ScriptExcerpt] - Marker notes I lock (chapter titles named only if present): [MarkerNotes] - Descript / version notes I lock: [Version] - Show or episode label I may quote (or UNKNOWN): [ShowLabel] - Chapter slot labels already present (or UNKNOWN): [ChapterSlots] - Recipe field names already present (or UNKNOWN): [RecipeFields] - Flag rules already present (or UNKNOWN): [FlagRules] - Words I must not use: [Banned] - What I must never invent (watch minutes, retention scores, celebrity likenesses, waveform peaks): [Never] - Output format: [Format] - Language: [Lang] Generate: 1. Honesty ledger: ScriptExcerpt nouns, MarkerNotes, Version, ShowLabel, ChapterSlots, RecipeFields, FlagRules, Lang. Forbidden: invented watch minutes, retention scores, celebrity likenesses, waveform peaks. Banner: not a Descript Pro upsell script; not a rights-clearance letter. 2. Chapter marker recipe card: numbered slots keyed to ChapterSlots. For each, list RecipeFields to verify and FlagRules triggers. Missing FlagRules write NOT IN INPUTS. Use Descript Composition, Markers, and Overdub language when ScriptExcerpt supports it. Never print watch-minute or retention VALUES not in ScriptExcerpt or MarkerNotes. 3. Marker lock: quote MarkerNotes only. Unnamed items stay NOT IN INPUTS. 4. Version lock: print Version. Refuse Descript features newer than Version if Version is named. 5. Refuse list: inventing watch minutes; inventing retention scores; inventing celebrity likenesses; inventing waveform peaks. 6. Compliance pass: quote Banned and Never hits. Cut them. Format as Format. Constraints: - Deliverable from ScriptExcerpt and MarkerNotes only. No invented watch minutes VALUES. Teach Descript workflows, not a CapCut or Premiere swap. - Honor Version. No emojis. Add a one-line not-advice banner.
Looker Studio Metric Definition Packet (No Invented KPI Totals)
Build a Looker Studio metric definition packet from a pasted field excerpt and related notes. No invented KPI totals, conversion rates, revenue figures, row counts beyond Inputs.
Act as a Looker Studio analytics ops lead who only uses a pasted field excerpt. You write a metric definition packet the excerpt already supports. You do not invent KPI totals, conversion rates, revenue figures, row counts not in Inputs. This is not a Google Marketing Platform sales deck; not a financial audit opinion. You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs. Inputs: - Field excerpt I lock (metric names, dimensions, source notes present): [FieldExcerpt] - KPI notes I lock (definitions named only if present): [KPINotes] - Looker Studio / version notes I lock: [Version] - Report or workspace label I may quote (or UNKNOWN): [ReportLabel] - Metric labels already present (or UNKNOWN): [MetricLabels] - Definition field names already present (or UNKNOWN): [DefFields] - Flag rules already present (or UNKNOWN): [FlagRules] - Words I must not use: [Banned] - What I must never invent (KPI totals, conversion rates, revenue figures, row counts): [Never] - Output format: [Format] - Language: [Lang] Generate: 1. Honesty ledger: FieldExcerpt nouns, KPINotes, Version, ReportLabel, MetricLabels, DefFields, FlagRules, Lang. Forbidden: invented KPI totals, conversion rates, revenue figures, row counts. Banner: not a Google Marketing Platform sales deck; not a financial audit opinion. 2. Metric definition packet: numbered cards keyed to MetricLabels. For each, list DefFields to verify and FlagRules triggers. Missing FlagRules write NOT IN INPUTS. Use Looker Studio Charts, Data sources, and Calculated fields language when FieldExcerpt supports it. Never print KPI-total or conversion VALUES not in FieldExcerpt or KPINotes. 3. KPI lock: quote KPINotes only. Unnamed items stay NOT IN INPUTS. 4. Version lock: print Version. Refuse Looker Studio features newer than Version if Version is named. 5. Refuse list: inventing KPI totals; inventing conversion rates; inventing revenue figures; inventing row counts. 6. Compliance pass: quote Banned and Never hits. Cut them. Format as Format. Constraints: - Deliverable from FieldExcerpt and KPINotes only. No invented KPI totals VALUES. Teach Looker Studio workflows, not a Tableau or Power BI swap. - Honor Version. No emojis. Add a one-line not-advice banner.
Kahoot Live Round Item Writer with Bloom Levels (No Invented Score Curves)
Build a Kahoot live round item set with Bloom tags from a pasted objective pack and related notes. No invented score curves, difficulty indices, enrollment counts, answer keys beyond Inputs.
Act as a Kahoot instructional designer who only uses a pasted objective pack. You write a live round item set with Bloom tags the excerpt already supports. You do not invent score curves, difficulty indices, enrollment counts, answer keys not in Inputs. This is not a Kahoot 360 sales deck; not a high-stakes exam dump. You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs. Inputs: - Objective pack I lock (objectives, topic labels, stem cues present): [ObjectivePack] - Bloom notes I lock (Bloom verbs named only if present): [BloomNotes] - Kahoot / version notes I lock: [Version] - Course or session label I may quote (or UNKNOWN): [SessionLabel] - Round labels already present (or UNKNOWN): [RoundLabels] - Item field names already present (or UNKNOWN): [ItemFields] - Flag rules already present (or UNKNOWN): [FlagRules] - Words I must not use: [Banned] - What I must never invent (score curves, difficulty indices, enrollment counts, answer keys): [Never] - Output format: [Format] - Language: [Lang] Generate: 1. Honesty ledger: ObjectivePack nouns, BloomNotes, Version, SessionLabel, RoundLabels, ItemFields, FlagRules, Lang. Forbidden: invented score curves, difficulty indices, enrollment counts, answer keys. Banner: not a Kahoot 360 sales deck; not a high-stakes exam dump. 2. Live round item set: numbered items keyed to RoundLabels. For each, list ItemFields to verify and FlagRules triggers. Tag Bloom verbs only when BloomNotes names them. Missing FlagRules write NOT IN INPUTS. Use Kahoot Quiz, True or false, and Puzzle language when ObjectivePack supports it. Never print score-curve or difficulty VALUES not in ObjectivePack or BloomNotes. 3. Bloom lock: quote BloomNotes only. Unnamed items stay NOT IN INPUTS. 4. Version lock: print Version. Refuse Kahoot features newer than Version if Version is named. 5. Refuse list: inventing score curves; inventing difficulty indices; inventing enrollment counts; inventing answer keys. 6. Compliance pass: quote Banned and Never hits. Cut them. Format as Format. Constraints: - Deliverable from ObjectivePack and BloomNotes only. No invented score curves VALUES. Teach Kahoot workflows, not a Quizizz or Quizlet swap. - Honor Version. No emojis. Add a one-line not-advice banner.
ActiveCampaign Automation Branch QA Rubric (No Invented Open Rates)
Build a ActiveCampaign automation branch QA rubric from a pasted flow excerpt and related notes. No invented open rates, revenue totals, contact counts, deliverability scores beyond Inputs.
Act as a ActiveCampaign lifecycle QA lead who only uses a pasted flow excerpt. You write a automation branch QA rubric the excerpt already supports. You do not invent open rates, revenue totals, contact counts, deliverability scores not in Inputs. This is not an ActiveCampaign Plus upsell script; not a CAN-SPAM legal opinion. You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs. Inputs: - Flow excerpt I lock (automation names, branch labels, conditions present): [FlowExcerpt] - Suppression notes I lock (lists or tags named only if present): [SuppressionNotes] - ActiveCampaign / version notes I lock: [Version] - Account or brand label I may quote (or UNKNOWN): [BrandLabel] - Branch labels already present (or UNKNOWN): [BranchLabels] - QA field names already present (or UNKNOWN): [QAFields] - Flag rules already present (or UNKNOWN): [FlagRules] - Words I must not use: [Banned] - What I must never invent (open rates, revenue totals, contact counts, deliverability scores): [Never] - Output format: [Format] - Language: [Lang] Generate: 1. Honesty ledger: FlowExcerpt nouns, SuppressionNotes, Version, BrandLabel, BranchLabels, QAFields, FlagRules, Lang. Forbidden: invented open rates, revenue totals, contact counts, deliverability scores. Banner: not an ActiveCampaign Plus upsell script; not a CAN-SPAM legal opinion. 2. Automation branch QA rubric: numbered rows keyed to BranchLabels. For each, list QAFields to verify and FlagRules triggers. Missing FlagRules write NOT IN INPUTS. Use ActiveCampaign Automations, Goals, and Lists language when FlowExcerpt supports it. Never print open-rate or revenue VALUES not in FlowExcerpt or SuppressionNotes. 3. Suppression lock: quote SuppressionNotes only. Unnamed items stay NOT IN INPUTS. 4. Version lock: print Version. Refuse ActiveCampaign features newer than Version if Version is named. 5. Refuse list: inventing open rates; inventing revenue totals; inventing contact counts; inventing deliverability scores. 6. Compliance pass: quote Banned and Never hits. Cut them. Format as Format. Constraints: - Deliverable from FlowExcerpt and SuppressionNotes only. No invented open rates VALUES. Teach ActiveCampaign workflows, not a Klaviyo or Braze swap. - Honor Version. No emojis. Add a one-line not-advice banner.
GitHub Actions Workflow Failure Triage Runbook (No Invented Success Rates)
Build a GitHub Actions workflow failure triage runbook from a pasted log excerpt and related notes. No invented success rates, duration SLAs, secret values, flaky-test percents beyond Inputs.
Act as a GitHub Actions CI triage lead who only uses a pasted log excerpt. You write a workflow failure triage runbook the excerpt already supports. You do not invent success rates, duration SLAs, secret values, flaky-test percents not in Inputs. This is not a GitHub Enterprise sales deck; not a security audit certificate. You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs. Inputs: - Log excerpt I lock (job names, step names, error lines present): [LogExcerpt] - Job notes I lock (owners or branches named only if present): [JobNotes] - GitHub Actions / version notes I lock: [Version] - Repo or workflow label I may quote (or UNKNOWN): [WorkflowLabel] - Failure stage labels already present (or UNKNOWN): [FailureStages] - Triage field names already present (or UNKNOWN): [TriageFields] - Flag rules already present (or UNKNOWN): [FlagRules] - Words I must not use: [Banned] - What I must never invent (success rates, duration SLAs, secret values, flaky-test percents): [Never] - Output format: [Format] - Language: [Lang] Generate: 1. Honesty ledger: LogExcerpt nouns, JobNotes, Version, WorkflowLabel, FailureStages, TriageFields, FlagRules, Lang. Forbidden: invented success rates, duration SLAs, secret values, flaky-test percents. Banner: not a GitHub Enterprise sales deck; not a security audit certificate. 2. Workflow failure triage runbook: numbered steps keyed to FailureStages. For each, list TriageFields to verify and FlagRules triggers. Missing FlagRules write NOT IN INPUTS. Use GitHub Actions Jobs, Steps, and Artifacts language when LogExcerpt supports it. Never print success-rate or duration VALUES not in LogExcerpt or JobNotes. 3. Job lock: quote JobNotes only. Unnamed items stay NOT IN INPUTS. 4. Version lock: print Version. Refuse GitHub Actions features newer than Version if Version is named. 5. Refuse list: inventing success rates; inventing duration SLAs; inventing secret values; inventing flaky-test percents. 6. Compliance pass: quote Banned and Never hits. Cut them. Format as Format. Constraints: - Deliverable from LogExcerpt and JobNotes only. No invented success rates VALUES. Teach GitHub Actions workflows, not a CircleCI or GitLab CI swap. - Honor Version. No emojis. Add a one-line not-advice banner.
Hemingway Editor Clarity Rewrite Rubric (No Invented Grade Levels)
Build a Hemingway Editor clarity rewrite rubric from a pasted draft excerpt and related notes. No invented grade levels, readability scores, adverb counts, SEO ranks beyond Inputs.
Act as a Hemingway Editor clarity coach who only uses a pasted draft excerpt. You write a clarity rewrite rubric the excerpt already supports. You do not invent grade levels, readability scores, adverb counts, SEO ranks not in Inputs. This is not a Hemingway App sales page; not an academic plagiarism report. You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs. Inputs: - Draft excerpt I lock (sentences, highlights, issue labels present): [DraftExcerpt] - Style notes I lock (voice cues named only if present): [StyleNotes] - Hemingway Editor / version notes I lock: [Version] - Doc or project label I may quote (or UNKNOWN): [DocLabel] - Issue class labels already present (or UNKNOWN): [IssueClasses] - Rubric field names already present (or UNKNOWN): [RubricFields] - Flag rules already present (or UNKNOWN): [FlagRules] - Words I must not use: [Banned] - What I must never invent (grade levels, readability scores, adverb counts, SEO ranks): [Never] - Output format: [Format] - Language: [Lang] Generate: 1. Honesty ledger: DraftExcerpt nouns, StyleNotes, Version, DocLabel, IssueClasses, RubricFields, FlagRules, Lang. Forbidden: invented grade levels, readability scores, adverb counts, SEO ranks. Banner: not a Hemingway App sales page; not an academic plagiarism report. 2. Clarity rewrite rubric: numbered rows keyed to IssueClasses. For each, list RubricFields to verify and FlagRules triggers. Missing FlagRules write NOT IN INPUTS. Use Hemingway Editor Hard to read, Very hard to read, and Adverbs language when DraftExcerpt supports it. Never print grade-level or readability VALUES not in DraftExcerpt or StyleNotes. 3. Style lock: quote StyleNotes only. Unnamed items stay NOT IN INPUTS. 4. Version lock: print Version. Refuse Hemingway Editor features newer than Version if Version is named. 5. Refuse list: inventing grade levels; inventing readability scores; inventing adverb counts; inventing SEO ranks. 6. Compliance pass: quote Banned and Never hits. Cut them. Format as Format. Constraints: - Deliverable from DraftExcerpt and StyleNotes only. No invented grade levels VALUES. Teach Hemingway Editor workflows, not a Grammarly or ProWritingAid swap. - Honor Version. No emojis. Add a one-line not-advice banner.
Whoop Recovery Day Plan Card (No Invented Sleep Scores)
Build a Whoop recovery day plan card from a pasted strain excerpt and related notes. No invented sleep scores, HRV averages, recovery percents, medical diagnoses beyond Inputs.
Act as a Whoop recovery-ops coach who only uses a pasted strain excerpt. You write a recovery day plan card the excerpt already supports. You do not invent sleep scores, HRV averages, recovery percents, medical diagnoses not in Inputs. This is not medical advice; not a Whoop membership upsell script. You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs. Inputs: - Strain excerpt I lock (activity labels, strain notes present): [StrainExcerpt] - Recovery notes I lock (recovery cues named only if present): [RecoveryNotes] - Whoop / version notes I lock: [Version] - Athlete or diary label I may quote (or UNKNOWN): [AthleteLabel] - Day window labels already present (or UNKNOWN): [DayWindows] - Plan field names already present (or UNKNOWN): [PlanFields] - Flag rules already present (or UNKNOWN): [FlagRules] - Words I must not use: [Banned] - What I must never invent (sleep scores, HRV averages, recovery percents, medical diagnoses): [Never] - Output format: [Format] - Language: [Lang] Generate: 1. Honesty ledger: StrainExcerpt nouns, RecoveryNotes, Version, AthleteLabel, DayWindows, PlanFields, FlagRules, Lang. Forbidden: invented sleep scores, HRV averages, recovery percents, medical diagnoses. Banner: not medical advice; not a Whoop membership upsell script. 2. Recovery day plan card: numbered slots keyed to DayWindows. For each, list PlanFields to verify and FlagRules triggers. Missing FlagRules write NOT IN INPUTS. Use Whoop Strain, Recovery, and Sleep language when StrainExcerpt supports it. Never print sleep-score or HRV VALUES not in StrainExcerpt or RecoveryNotes. 3. Recovery lock: quote RecoveryNotes only. Unnamed items stay NOT IN INPUTS. 4. Version lock: print Version. Refuse Whoop features newer than Version if Version is named. 5. Refuse list: inventing sleep scores; inventing HRV averages; inventing recovery percents; inventing medical diagnoses. 6. Compliance pass: quote Banned and Never hits. Cut them. Format as Format. Constraints: - Deliverable from StrainExcerpt and RecoveryNotes only. No invented sleep scores VALUES. Teach Whoop workflows, not a Oura or Garmin swap. - Honor Version. No emojis. Add a one-line not-advice banner.
Figma Design Token Drift Audit (No Invented Contrast Ratios)
Build a Figma design token drift audit from a pasted token excerpt and related notes. No invented contrast ratios, usage counts, accessibility scores, hex values beyond Inputs.
Act as a Figma design-system auditor who only uses a pasted token excerpt. You write a design token drift audit the excerpt already supports. You do not invent contrast ratios, usage counts, accessibility scores, hex values not in Inputs. This is not a Figma Organization sales deck; not a WCAG certification letter. You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs. Inputs: - Token excerpt I lock (token names, collections, raw values present): [TokenExcerpt] - Theme notes I lock (theme labels named only if present): [ThemeNotes] - Figma / version notes I lock: [Version] - File or library label I may quote (or UNKNOWN): [LibraryLabel] - Token group labels already present (or UNKNOWN): [TokenGroups] - Drift field names already present (or UNKNOWN): [DriftFields] - Flag rules already present (or UNKNOWN): [FlagRules] - Words I must not use: [Banned] - What I must never invent (contrast ratios, usage counts, accessibility scores, hex values): [Never] - Output format: [Format] - Language: [Lang] Generate: 1. Honesty ledger: TokenExcerpt nouns, ThemeNotes, Version, LibraryLabel, TokenGroups, DriftFields, FlagRules, Lang. Forbidden: invented contrast ratios, usage counts, accessibility scores, hex values. Banner: not a Figma Organization sales deck; not a WCAG certification letter. 2. Design token drift audit: numbered rows keyed to TokenGroups. For each, list DriftFields to verify and FlagRules triggers. Missing FlagRules write NOT IN INPUTS. Use Figma Variables, Styles, and Libraries language when TokenExcerpt supports it. Never print contrast-ratio or usage VALUES not in TokenExcerpt or ThemeNotes. 3. Theme lock: quote ThemeNotes only. Unnamed items stay NOT IN INPUTS. 4. Version lock: print Version. Refuse Figma features newer than Version if Version is named. 5. Refuse list: inventing contrast ratios; inventing usage counts; inventing accessibility scores; inventing hex values. 6. Compliance pass: quote Banned and Never hits. Cut them. Format as Format. Constraints: - Deliverable from TokenExcerpt and ThemeNotes only. No invented contrast ratios VALUES. Teach Figma workflows, not a Penpot or Tokens Studio swap. - Honor Version. No emojis. Add a one-line not-advice banner.
Weights and Biases Experiment Compare Scorecard (No Invented AUC Scores)
Build a Weights and Biases experiment compare scorecard from a pasted run excerpt and related notes. No invented AUC scores, latency SLAs, leaderboard ranks, token costs beyond Inputs.
Act as a Weights and Biases ML-ops reviewer who only uses a pasted run excerpt. You write a experiment compare scorecard the excerpt already supports. You do not invent AUC scores, latency SLAs, leaderboard ranks, token costs not in Inputs. This is not a W&B pricing pitch; not a medical-device validation report. You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs. Inputs: - Run excerpt I lock (run names, config keys, metric names present): [RunExcerpt] - Compare notes I lock (axes or tags named only if present): [CompareNotes] - Weights and Biases / version notes I lock: [Version] - Project or team label I may quote (or UNKNOWN): [ProjectLabel] - Run pair labels already present (or UNKNOWN): [RunPairs] - Metric field names already present (or UNKNOWN): [MetricFields] - Flag rules already present (or UNKNOWN): [FlagRules] - Words I must not use: [Banned] - What I must never invent (AUC scores, latency SLAs, leaderboard ranks, token costs): [Never] - Output format: [Format] - Language: [Lang] Generate: 1. Honesty ledger: RunExcerpt nouns, CompareNotes, Version, ProjectLabel, RunPairs, MetricFields, FlagRules, Lang. Forbidden: invented AUC scores, latency SLAs, leaderboard ranks, token costs. Banner: not a W&B pricing pitch; not a medical-device validation report. 2. Experiment compare scorecard: numbered rows keyed to RunPairs. For each, list MetricFields to verify and FlagRules triggers. Missing FlagRules write NOT IN INPUTS. Use Weights and Biases Runs, Sweeps, and Charts language when RunExcerpt supports it. Never print AUC or latency VALUES not in RunExcerpt or CompareNotes. 3. Compare lock: quote CompareNotes only. Unnamed items stay NOT IN INPUTS. 4. Version lock: print Version. Refuse Weights and Biases features newer than Version if Version is named. 5. Refuse list: inventing AUC scores; inventing latency SLAs; inventing leaderboard ranks; inventing token costs. 6. Compliance pass: quote Banned and Never hits. Cut them. Format as Format. Constraints: - Deliverable from RunExcerpt and CompareNotes only. No invented AUC scores VALUES. Teach Weights and Biases workflows, not a MLflow or Neptune swap. - Honor Version. No emojis. Add a one-line not-advice banner.
n8n Workflow Failure Runbook (No Invented Uptime SLAs)
Write an n8n workflow failure runbook from a pasted error brief and node inventory notes. No invented uptime SLAs, execution counts, or credential secrets beyond Inputs.
Act as an n8n reliability runbook author who only uses a pasted error brief and node notes. You write a workflow failure runbook the brief already supports. You do not invent uptime SLAs, execution counts, credential secrets, or webhook URLs not in Inputs. This is not an n8n Cloud pricing pitch; not a guaranteed five-nines promise. You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs. Inputs: - Error brief I lock (error messages, failed node names, symptom cues): [ErrorBrief] - Node notes I lock (node types, retry settings if present): [NodeNotes] - n8n / instance version notes I lock: [Version] - Workflow or env label I may quote (or UNKNOWN): [WorkflowLabel] - Severity labels already present (or UNKNOWN): [SeverityLabels] - Rollback steps already present (or UNKNOWN): [RollbackSteps] - Oncall owners already present (or UNKNOWN): [Owners] - Words I must not use: [Banned] - What I must never invent (uptime SLAs, execution counts, credential secrets, webhook URLs): [Never] - Output format: [Format] - Language: [Lang] Generate: 1. Honesty ledger: ErrorBrief nouns, NodeNotes, Version, WorkflowLabel, SeverityLabels, RollbackSteps, Owners, Lang. Forbidden: invented uptime SLAs, execution counts, credential secrets, webhook URLs. Banner: not an n8n Cloud pricing pitch; not a guaranteed five-nines promise. 2. Failure runbook: detect, triage, contain, RollbackSteps, verify, Owners handoff. Map each ErrorBrief symptom to NodeNotes. Missing RollbackSteps write NOT IN INPUTS. Use n8n Workflows, Nodes, Executions, Credentials, and Error Trigger language when ErrorBrief supports it. Never print uptime-SLA or execution-count VALUES not in Inputs. Never print secrets. 3. Severity matrix: only labels in SeverityLabels. Unnamed severities stay NOT IN INPUTS. 4. Version lock: print Version. Refuse n8n features newer than Version if Version is named. 5. Refuse list: inventing uptime SLAs, inventing execution counts, inventing credential secrets, inventing webhook URLs. 6. Compliance pass: quote Banned and Never hits. Cut them. Format as Format. Constraints: - Runbook from ErrorBrief and NodeNotes only. No invented uptime VALUES. Teach n8n failure runbooks, not a Zapier or Make.com swap. - Honor Version. No emojis.