30 GPT-5 Prompts for Long-Context Work (1M Tokens, Copy-Paste)

You are a senior software engineer conducting a full codebase audit.

I am pasting the complete contents of my repository below. 
Each file is preceded by a header in the format: === FILE: [path/to/file] ===

[PASTE FULL CODEBASE HERE]

Perform a structured audit covering:

1. Architecture Overview
   - How is the codebase organized? What are the main modules and their responsibilities?
   - What architectural patterns are used (MVC, event-driven, layered, etc.)?
   - Are there clear boundaries between layers? If not, which files are the worst offenders — quote them by name.

2. Code Quality Issues
   - List every file with critical quality problems (duplicated logic, god objects, 
     deep nesting, inconsistent error handling).
   - For each issue: quote the file name and relevant lines verbatim, explain the problem, 
     and recommend the fix.

3. Security Concerns
   - Identify any hardcoded secrets, unsafe inputs, SQL injection risks, or auth gaps.
   - Quote the exact file and line. Do not infer — if you do not see an issue, say so explicitly.

4. Dependency Health
   - List all dependencies referenced across the codebase.
   - Flag any that appear unused, duplicated, or potentially vulnerable based on their pattern of use.

5. Priority Refactor List
   - Rank the top 5 changes by impact. For each: file name, the problem, and a one-paragraph fix plan.

Output format: Use ## headers for each section. Use a markdown table for the Priority Refactor List 
(columns: Rank | File | Problem | Recommended Fix).

If any section has no findings, say "No issues found in this category" — do not skip the section.

You are a security engineer. I am providing the full source code of my application below.
Each file is labeled with === FILE: [path] ===

[PASTE FULL CODEBASE HERE]

Perform a security scan. For every finding, you must:
- Quote the exact file path and relevant code verbatim
- Identify the vulnerability class (e.g., XSS, SSRF, insecure deserialization, hardcoded secret)
- Rate severity: CRITICAL / HIGH / MEDIUM / LOW
- Explain the attack vector in one sentence
- Provide a concrete fix (code snippet preferred)

Scan for:
1. Hardcoded credentials, API keys, tokens, or passwords
2. SQL injection or NoSQL injection risks
3. User input passed to shell commands, file paths, or eval()
4. Authentication or authorization bypasses
5. Insecure direct object references
6. Missing rate limiting or brute-force protections
7. Sensitive data logged or exposed in error messages
8. Unsafe deserialization or prototype pollution

Output as a markdown table with columns: Severity | File | Line(s) | Vulnerability | Fix Summary
Then provide full remediation detail for every CRITICAL and HIGH finding below the table.

If a category has no findings, state that explicitly. Do not omit categories.

You are a staff engineer performing a dead-code audit.

Below is the complete codebase. Each file is labeled === FILE: [path] ===

[PASTE FULL CODEBASE HERE]

Identify all dead code — functions, classes, variables, routes, exports, and imported modules 
that are defined but never called or referenced anywhere in the codebase.

For each item:
1. Quote the file name and the dead code verbatim (first 3 lines minimum)
2. Confirm it is unreachable by showing that no other file references it 
   (quote the search scope — "searched all [N] files, no references found")
3. Assess safe-to-delete confidence: HIGH (clearly unused), MEDIUM (may be called dynamically), 
   LOW (used in a non-obvious way — explain)

Do not infer usage — if a function is called dynamically (e.g., via reflection or string lookup) 
and you cannot confirm it is dead, say so explicitly and mark confidence LOW.

Output format:
- Summary table: File | Symbol | Type | Confidence
- Full findings section with quotes and evidence below the table
- Estimated lines-of-code reduction if all HIGH-confidence items are deleted

You are a principal engineer. I need a refactor proposal for the module(s) below.

Context: [DESCRIBE THE MODULE — e.g., "authentication service, ~800 lines across 4 files"]
Goal: [WHAT YOU WANT — e.g., "reduce coupling, improve testability, split into smaller units"]

Below is the full source. Each file is labeled === FILE: [path] ===

[PASTE MODULE FILES HERE]

Produce a detailed refactor proposal:

1. Current State Analysis
   - What does this module do?
   - What are the top 3 structural problems? Quote the specific code that demonstrates each problem.

2. Proposed Structure
   - New file breakdown: what files to create, rename, or delete
   - Responsibility of each new unit (one sentence per file)
   - Dependency diagram (describe in text — what depends on what)

3. Migration Steps
   - Ordered, numbered steps to move from current state to proposed state
   - Each step should be independently committable (no "big bang" refactors)
   - Note which steps carry regression risk and how to test around them

4. Effort Estimate
   - Lines of code affected (approximate)
   - Estimated developer-hours per step

If the current structure is already clean in some area, say so — do not fabricate problems.

You are a senior architect. I am providing a full codebase below.
Each file is labeled === FILE: [path] ===

[PASTE FULL CODEBASE HERE]

Generate a complete dependency map of the codebase:

1. Module Dependency Graph
   - List every module/file and what it imports (quote the import statements verbatim)
   - Identify circular dependencies — for each one, quote both files and the circular reference chain
   - Identify tightly coupled modules (modules that import from many others)

2. External Dependency Inventory
   - List every third-party package imported across the entire codebase
   - Count how many files reference each package
   - Flag packages imported in only one file (candidate for removal or inlining)

3. Entry Points
   - What are the entry points of the application? 
   - Trace the execution path from the primary entry point to the 3 most critical modules.
     Quote the function calls at each step.

4. Coupling Score
   - For each module, calculate: (number of modules it imports) + (number of modules that import it)
   - Rank top 10 most-coupled modules in a table: Module | Imports | Imported-By | Coupling Score

Output format: Use ## headers. Use markdown tables where specified.
If information is not determinable from the source (e.g., dynamic requires), say so explicitly.

You are a contract analyst. I am providing two versions of the same contract below.

CONTRACT VERSION A (label all references as "V-A"):
[PASTE VERSION A]

CONTRACT VERSION B (label all references as "V-B"):
[PASTE VERSION B]

Compare the two versions and produce:

1. Changed Clauses Table
   Columns: Clause/Section | V-A Text (verbatim) | V-B Text (verbatim) | Change Type | Risk Impact
   Change types: Added | Removed | Modified | Renumbered
   Risk impact: Higher Risk | Lower Risk | Neutral | Unclear

2. New Obligations in V-B
   - List every obligation that exists in V-B but not V-A
   - Quote the clause verbatim from V-B with its section identifier

3. Removed Protections in V-B
   - List every protective clause present in V-A that was removed or weakened in V-B
   - Quote both versions verbatim for each

4. Executive Summary
   - 5-bullet summary of the most significant changes, ordered by risk impact

Do not summarize without quoting. If a section appears identical in both versions, 
confirm that explicitly — do not skip it.
If information is not present in either version, say so explicitly — do not infer.

You are a contract risk analyst. Review the full contract below and extract every clause 
that poses a risk to [PARTY — e.g., "the buyer" / "the service provider" / "our company"].

[PASTE FULL CONTRACT HERE]

For every risk clause:
1. Quote the clause verbatim, including the section number
2. Categorize the risk: Financial | Liability | IP Ownership | Termination | 
   Confidentiality | Non-Compete | Indemnification | Other
3. Rate severity: HIGH (existential/major financial exposure) | 
   MEDIUM (significant but manageable) | LOW (minor)
4. Explain the risk in plain English (2 sentences maximum)
5. Recommend: Accept As-Is | Negotiate Modification | Remove | Add Protective Clause

Output as a markdown table with the columns above, followed by full quoted text 
for all HIGH-severity items.

Important: Only flag clauses that actually appear in the document. 
If a common risk clause (e.g., limitation of liability) is absent, 
note its absence in a separate "Missing Protective Clauses" section at the end.
Do not infer risk from clauses that are not present.

You are a contract attorney assistant. Review the contract below and generate a redline 
— a list of proposed modifications with legal justification.

Representing party: [WHO YOU REPRESENT — e.g., "the vendor" / "the licensee"]
Priority concerns: [LIST TOP 3 — e.g., "liability cap, IP ownership, termination for convenience"]

[PASTE FULL CONTRACT HERE]

For each proposed change:
1. Quote the original clause verbatim with its section number
2. Provide the proposed replacement language
3. Justify the change in 2-3 sentences — what risk it reduces or what right it preserves
4. Classify urgency: Must-Have | Strong Preference | Nice-to-Have

Format output as:
## Section [X.X] — [Clause Name]
**Original:** [quoted text]
**Proposed:** [replacement text]
**Justification:** [explanation]
**Urgency:** [classification]

Cover every section of the document. If a section requires no changes, 
write "No changes recommended — clause is standard and acceptable."
Do not skip sections. Do not fabricate clauses that do not exist in the document.

You are a contract compliance analyst. Extract every obligation from the contract below 
and build an obligation matrix.

[PASTE FULL CONTRACT HERE]

An obligation is any clause that requires a party to DO something, 
REFRAIN from something, or DELIVER something by a date or condition.

For every obligation found:
- Quote the source clause verbatim with section number
- Identify the obligated party
- Describe the obligation in plain English (one sentence)
- Identify the trigger (effective date, milestone, event, ongoing)
- Identify the deadline or frequency if stated — if not stated, write "Not specified"
- Flag any penalty or consequence for non-compliance — quote verbatim if present

Output as a markdown table:
Section | Obligated Party | Obligation Summary | Trigger | Deadline/Frequency | Penalty if Any

After the table, list any obligations where the deadline or responsible party 
is ambiguous — quote the clause and explain the ambiguity.

Do not infer obligations not stated in the document.

You are a contract manager. I need a formal change log documenting every difference 
between the following two contract versions.

VERSION 1 (original, dated [DATE]):
[PASTE VERSION 1]

VERSION 2 (revised, dated [DATE]):
[PASTE VERSION 2]

Produce a structured change log:

1. Change Log Table
   Columns: Section | Change Type | Original Language (verbatim) | Revised Language (verbatim) | 
   Business Impact
   Change types: Addition | Deletion | Substitution | Renumbering | Formatting Only

2. Net Risk Movement
   - Has the contract become more or less favorable to [PARTY]? 
   - Support your answer by quoting the 3 most impactful changes verbatim.

3. Unresolved Items
   - List any bracketed placeholders, TBDs, or blank fields in Version 2 that require 
     completion before execution. Quote each one verbatim with its location.

Present every change — no matter how minor. 
For changes you classify as "Formatting Only," confirm that the legal meaning 
is identical between versions. Do not skip any section.

You are a market research analyst. I am providing [NUMBER] industry reports below.
Each report is preceded by a header: === REPORT: [Company/Source Name] ===

[PASTE ALL REPORTS HERE]

Synthesize a competitive landscape analysis covering:

1. Market Positioning Map
   - For each company/product mentioned across the reports, identify:
     a) Their stated positioning (quote the source report and relevant passage verbatim)
     b) Their key differentiators
     c) Their apparent target customer
   - Present as a table: Company | Positioning | Differentiators | Target Customer | Source

2. Competitive Claims Cross-Reference
   - Where do two or more reports make competing or contradictory claims about the same company 
     or market? Quote both passages verbatim and identify the conflict.

3. White Space
   - What needs or segments are described as underserved across multiple reports? 
   - Quote the evidence from each report that supports each white space identified.

4. Consensus Trends
   - What trends appear in 3 or more reports? List each trend and quote one representative 
     passage per report that confirms it.

If a company or topic is mentioned in only one report, note that — do not present single-source 
findings as consensus. If information is absent from the reports, say so explicitly.

You are a research synthesist. I am providing the full text of [NUMBER] research papers below.
Each paper is preceded by: === PAPER: [Title — Authors — Year] ===

[PASTE ALL PAPERS HERE]

Produce a structured synthesis:

1. Research Landscape
   - What is the primary question or domain these papers collectively address?
   - Group papers into 3-5 thematic clusters. Name each cluster and list which papers belong to it, 
     citing the paper title and year.

2. Key Findings Table
   Columns: Paper (Title + Year) | Core Finding (verbatim quote from abstract or conclusion) | 
   Methodology | Sample Size / Scope | Strength of Evidence

3. Points of Consensus
   - What findings are supported by 3 or more papers? 
   - For each point of consensus, quote one passage per supporting paper verbatim 
     and cite title + year.

4. Points of Contradiction
   - Where do papers reach conflicting conclusions? 
   - Quote the conflicting passages verbatim, identify the papers, and propose a hypothesis 
     for why the findings differ (methodology, population, time period).

5. Research Gaps
   - What questions do the authors themselves identify as unresolved? 
   - Quote the specific passages (including paper title and year) where gaps are stated.

Do not synthesize findings not present in the provided papers. 
If a paper's methodology or conclusions are unclear from the provided text, say so.

You are a fact-extraction analyst. I am providing a corpus of documents below.
Each document is labeled === DOCUMENT: [Title or ID] ===

[PASTE ALL DOCUMENTS HERE]

Task: Extract every factual claim related to [TOPIC — e.g., "AI adoption in healthcare" / 
"interest rate effects on housing" / "carbon capture efficiency"].

For every claim extracted:
1. Quote the claim verbatim (do not paraphrase)
2. Cite the source document name/ID and the surrounding context (2 sentences before and after)
3. Classify the claim type: 
   Statistical | Causal | Predictive | Definitional | Anecdotal | Expert Opinion
4. Note if the claim is qualified (e.g., "according to X," "in some cases," "preliminary data suggests")

Output format: numbered list.
After the list, group claims by type in a summary table:
Claim # | Type | Source Document | Key Qualifier (if any)

Important: Only extract claims that explicitly appear in the documents. 
Do not add external knowledge. If no relevant claims are found in a document, 
note that document explicitly at the end.

You are a research integrity analyst. I am providing multiple documents below.
Each document is labeled === DOCUMENT: [ID or Title] ===

[PASTE ALL DOCUMENTS HERE]

Find every internal and cross-document contradiction — any place where 
two passages make mutually exclusive or conflicting factual claims.

For every contradiction found:
1. Quote both conflicting passages verbatim
2. Identify the source document and location for each
3. Describe the contradiction precisely: what specifically conflicts (a number, a date, 
   a causal claim, a factual assertion)
4. Classify the conflict type:
   - Direct Contradiction (statements are mutually exclusive)
   - Numerical Discrepancy (different figures for the same metric)
   - Temporal Conflict (different dates or sequences for the same event)
   - Definitional Conflict (same term used differently)
   - Scope Conflict (one document qualifies a claim the other makes absolute)

Rank contradictions by significance: HIGH (changes meaning materially) | MEDIUM | LOW

Do not flag nuance or complexity as contradiction. 
Only flag where two statements cannot both be true as written.
If no contradictions exist between a given pair of documents, state that explicitly.

You are a research analyst. I will ask a question, and you will answer it 
using only the documents I provide. You are not permitted to use outside knowledge.

Question: [YOUR RESEARCH QUESTION]

Documents:
[PASTE ALL DOCUMENTS HERE — each labeled === DOCUMENT: [ID] ===]

Answer the question using the following structure:

1. Direct Answer
   State your answer in 1-3 sentences. If the documents do not contain sufficient evidence 
   to answer the question, say so explicitly — do not speculate.

2. Supporting Evidence
   For each piece of evidence supporting the answer:
   - Quote the relevant passage verbatim
   - Cite the document ID and contextual location
   - Rate evidence strength: Strong (direct, specific data) | Moderate (indirect or qualified) | 
     Weak (anecdotal or single-source)

3. Contradicting Evidence
   List any passages from the documents that complicate or contradict the answer.
   Quote verbatim. Do not omit unfavorable evidence.

4. Confidence Level
   Rate overall confidence: HIGH | MEDIUM | LOW
   Justify based on the quantity and quality of evidence above.

5. Information Gaps
   What additional information would be needed to answer this question with higher confidence? 
   Be specific — name the type of data or source.

You are a project manager assistant. I am providing transcripts from [NUMBER] meetings below.
Each transcript is labeled === MEETING: [Date — Title or Project Name] ===

[PASTE ALL TRANSCRIPTS HERE]

Produce a consolidated decision log across all meetings:

1. Decision Registry
   Table columns: Decision # | Date | Meeting | Decision Made (verbatim or close paraphrase) | 
   Decision Maker(s) | Conditions or Caveats | Was It Reversed Later?

   For "Was It Reversed Later?" — scan all subsequent transcripts and note if the decision 
   was revisited, changed, or abandoned. Quote the reversal passage if found.

2. Decisions Still Open
   List every item that was framed as a pending decision or "to be decided" across the transcripts.
   Quote the relevant passage and the meeting it appeared in.

3. Decision Velocity
   Which meetings produced the most decisions? Rank by count.
   Which topics generated the most decision-making discussion (measured by how many meetings 
   touched the same topic)?

Quote all decisions from transcript text. Do not invent decisions.
If a meeting transcript is incomplete or unclear, note that.

You are a qualitative analyst. I am providing meeting or conversation transcripts 
spanning [TIME PERIOD] below. Each transcript is labeled === TRANSCRIPT: [Date] ===

[PASTE ALL TRANSCRIPTS HERE]

Track sentiment toward [TOPIC OR PROJECT — e.g., "the product roadmap" / "the partnership" / 
"the budget situation"] across transcripts over time.

For each transcript:
1. Overall Sentiment Toward [TOPIC]: Positive | Neutral | Mixed | Negative | Critical
2. Key sentiment indicators — quote 2-3 passages verbatim that drove your rating
3. Any sentiment shift within the same transcript (note who drove the shift and quote them)

Then:
4. Sentiment Timeline Table
   Columns: Date | Overall Sentiment | Key Quote (verbatim) | Who Was Most Positive | 
   Who Was Most Negative

5. Inflection Points
   Identify any transcripts where sentiment shifted significantly compared to the previous one.
   Quote the passages that mark the shift. Propose a likely cause based only on what 
   is stated in the transcripts — do not speculate beyond the text.

If sentiment toward [TOPIC] is not clearly expressed in a transcript, 
note that rather than inferring a rating.

You are an organizational dynamics analyst. I am providing meeting transcripts below.
Each is labeled === MEETING: [Date — Attendees] ===

[PASTE ALL TRANSCRIPTS HERE]

Identify signals of unspoken tension, conflict avoidance, or interpersonal friction.

Look for:
- Hedge language following another person's proposal ("that's interesting, but..." / 
  "I wonder if we've considered...")
- Sudden topic changes when a sensitive subject is raised
- Questions framed as challenges ("are we sure that's the right approach?")
- Absent stakeholders mentioned but not present
- Decisions made without apparent consensus
- Follow-up actions assigned to specific individuals that seem disproportionate

For every finding:
1. Quote the transcript passage verbatim (include speaker label if present)
2. Note the meeting date and approximate location in the transcript
3. Describe what the tension signal suggests — in one sentence, not a paragraph
4. Rate confidence: HIGH (explicit friction) | MEDIUM (pattern is suggestive) | 
   LOW (possible but ambiguous)

Do not diagnose interpersonal relationships or make character assessments.
Only describe what the transcript text supports. 
For every LOW-confidence finding, state why it might have an innocent explanation.

You are a project historian. I am providing all transcripts, notes, and documents 
related to [PROJECT NAME] below. Each file is labeled === SOURCE: [Type — Date] ===

[PASTE ALL SOURCE MATERIAL HERE]

Reconstruct the complete history of this project from the materials provided.

1. Project Timeline
   Table columns: Date | Event | Source (file label + verbatim quote that confirms the event)

   Include: kickoff decisions, scope changes, budget events, personnel changes, 
   milestone completions or failures, external dependencies, and any pivots.

2. Key Inflection Points
   Identify the 5 most consequential moments in the project history.
   For each: describe the moment, quote the primary evidence, and explain the downstream impact 
   (based only on what subsequent documents show).

3. What Was Planned vs What Happened
   Where do the documents show a gap between original plans and actual execution?
   Quote both the original plan and the evidence of deviation verbatim.

4. Open Threads
   What issues, risks, or action items appear in the documents with no documented resolution?
   List each and quote the source.

Limit your reconstruction to what is documented. 
Where the record is silent, say so — do not fill gaps with inference.

You are a customer insights analyst. I am providing transcripts from [NUMBER] customer calls below.
Each transcript is labeled === CALL: [Date — Customer Segment if known] ===

[PASTE ALL TRANSCRIPTS HERE]

Extract themes from across all customer calls:

1. Pain Points Mentioned
   - List every distinct pain point raised by customers
   - For each pain point: count how many calls mentioned it, 
     quote one representative customer statement verbatim per unique call that raised it,
     and note which calls mentioned it (by label)

2. Feature Requests and Wants
   - Same structure: list, count, quote, attribute

3. Competitor Mentions
   - Every time a competitor or alternative solution is mentioned by a customer:
     quote verbatim, note the call, and summarize the context (positive, negative, comparison)

4. Sentiment by Segment
   - If call labels include segment information, break down sentiment by segment
   - If not, note that segmentation is unavailable

5. Most Urgent Issues
   - Rank the top 5 themes by: frequency × intensity of language
   - For "intensity," quote the strongest language a customer used on that topic

Present only what customers actually said. Do not editorialize or add product recommendations.
If a theme appears in only one call, label it as single-source.

You are a financial analyst. I am providing the full text of [COMPANY]'s 
[FILING TYPE — e.g., 10-K / 10-Q / annual report] for [FISCAL YEAR] below.

[PASTE FULL FILING HERE]

Conduct a structured analysis:

1. Key Financial Metrics
   Extract the following from the filing — quote the source table or section verbatim for each:
   - Revenue (total and by segment if available)
   - Gross margin
   - Operating income / loss
   - Net income / loss
   - Free cash flow
   - Cash and equivalents
   - Total debt
   - Any forward guidance figures stated

2. Risk Factors
   List the top 10 risk factors as stated in the filing (quote each one verbatim from the 
   Risk Factors section). For each, note if this risk factor is new compared to [PRIOR PERIOD] 
   — only flag as new if the filing itself indicates this.

3. Management Commentary
   Quote the 5 most significant statements from the MD&A section that signal 
   strategic direction, concern, or opportunity. For each, explain in one sentence 
   what it signals.

4. Red Flags
   Identify any disclosures that warrant closer scrutiny: 
   going concern language, restatements, material weaknesses, litigation, 
   related-party transactions. Quote each verbatim with section reference.

If a metric is not present in the filing, state that explicitly — do not estimate.

You are a financial analyst. I am providing quarterly reports for [COMPANY] 
covering [NUMBER] quarters below. Each report is labeled === Q[X] [YEAR] ===

[PASTE ALL QUARTERLY REPORTS HERE]

Extract trends across quarters:

1. Metric Trend Table
   Rows: [one row per quarter, chronological]
   Columns: Quarter | Revenue | Gross Margin % | Operating Income | Net Income | 
   Cash Flow from Operations | [ADD ANY SEGMENT-SPECIFIC METRICS]
   Source: for each cell, cite the exact section and table name in the corresponding report.

2. Quarter-over-Quarter Changes
   For each consecutive pair of quarters, calculate the change in each metric.
   Flag any quarter where a metric moved more than [THRESHOLD — e.g., 10%] 
   and quote the management commentary that addressed it (if present).

3. Stated vs Actual
   Where did management provide guidance in one quarter and report results in the next?
   Table: Quarter | Guidance Metric | Guided Value (verbatim) | Actual Value | Beat/Miss

4. Narrative Shifts
   Has management language around [TOPIC — e.g., growth outlook / margin pressure / 
   a specific product line] changed in tone across quarters?
   Quote one representative passage per quarter on that topic.

Do not calculate or infer figures not explicitly stated in the filings.
Flag any quarters where the requested metric was not disclosed.

You are a research director. I am providing analyst reports on [COMPANY or SECTOR] 
from [NUMBER] different firms below. Each report is labeled === ANALYST: [Firm Name — Date] ===

[PASTE ALL ANALYST REPORTS HERE]

Synthesize across all reports:

1. Ratings and Price Targets
   Table: Firm | Rating | Price Target | Date | Change from Prior (if stated)

2. Bull Case vs Bear Case
   - What are the shared bull arguments? Quote one passage per report that contributes.
   - What are the shared bear arguments? Quote one passage per report that contributes.
   - Where do analysts disagree? Quote the conflicting positions and name the firms.

3. Key Assumptions Comparison
   Extract the core model assumptions each analyst uses (revenue growth rate, margin expansion, 
   multiple used). Present in a table. If an analyst does not state an assumption explicitly, 
   write "Not stated" — do not infer.

4. Unique Insights
   What does each report say that no other report says? 
   Quote the unique insight and name the firm.

5. Consensus Summary
   What is the median rating and price target? 
   What is the single most common reason cited for or against the stock? 
   Quote representative text from at least 3 reports.

You are a documentation specialist. I am providing the complete contents of a 
knowledge base or documentation corpus below. Each document is labeled === DOC: [Title] ===

[PASTE FULL KNOWLEDGE BASE HERE]

Build a comprehensive FAQ from this knowledge base:

1. Extract every question that is either:
   a) Explicitly answered in the documentation (common question formats, "how do I," 
      "what is," "why does," "when should")
   b) Implied by the topic coverage (a topic that is explained signals a user question)

2. For each FAQ entry:
   - State the question in natural language as a user would ask it
   - Provide the answer drawn directly from the documentation 
     (quote the source passage verbatim, then paraphrase into a clean answer)
   - Cite the source document label and section

3. Categorize the FAQ:
   - Group questions into logical categories
   - Within each category, order from basic to advanced

4. Coverage Gaps
   - Identify topics or terms that appear in the documentation but are never fully explained.
   - List each gap and quote the passage where the unexplained term or topic appears.

Answers must be grounded in the provided documentation only. 
If a question cannot be answered from the provided content, write "Not covered in current documentation" and list it in the Gaps section.

You are a data integrity analyst. I am providing a large dataset or report corpus below.
[DESCRIBE THE DATA FORMAT — e.g., "CSV with columns: Date, Product, Region, Revenue, Units"]

[PASTE DATA OR REPORT HERE]

Hunt for anomalies using the following criteria:

1. Numerical Anomalies
   - Values that are statistical outliers (more than 3x the surrounding average for that metric)
   - Sudden discontinuities between consecutive periods
   - Zero or negative values in fields that should always be positive
   - Quote the exact row/entry and value for each anomaly found

2. Logical Inconsistencies
   - Entries where two fields contradict each other 
     (e.g., units sold = 0 but revenue > 0)
   - Quote the exact entry verbatim

3. Missing Data Patterns
   - Are there systematic gaps? (e.g., one region never reports, one product category 
     appears only in some periods)
   - Note the pattern and the extent of the gap

4. Repeated or Duplicate Entries
   - Identify exact or near-exact duplicate rows
   - Quote both entries and note the location of each

5. Anomaly Summary Table
   Columns: Type | Location (row/section) | Description | Severity | Recommended Action

If no anomalies are found in a category, state that explicitly. 
Do not flag normal variation as anomalous. Provide evidence for every finding.

You are a conversation analyst. I am providing a long conversation thread or message history below.
[DESCRIBE THE CONTEXT — e.g., "Slack thread about product roadmap planning, 
spanning 3 weeks, ~200 messages"]

[PASTE FULL CONVERSATION THREAD HERE]

Produce a structured summary:

1. Executive Summary
   What is this conversation about? What was resolved and what remains open?
   Write 3-5 sentences. Every sentence must be supported by a quoted passage below.

2. Timeline of Key Developments
   List the 10 most significant moments in this conversation, in chronological order.
   For each: quote the message verbatim, note the sender (if labeled), and note the date/timestamp.

3. Decisions Made
   List every decision that was reached in this thread.
   Quote the message that confirmed the decision.

4. Action Items
   List every task, assignment, or follow-up that was agreed upon.
   Format: [Owner if stated] will [action] — quote the source message verbatim.

5. Open Questions
   What was raised but never resolved? Quote each unresolved item verbatim.

If a message is ambiguous about whether a decision was made or merely discussed, 
note the ambiguity and quote the passage.

You are a conversation memory assistant. I am providing a long conversation history below.

[PASTE FULL CONVERSATION HISTORY HERE]

I will ask you a series of questions about specific things said in this conversation.
For each question, your task is:

1. Search the full conversation for every relevant passage
2. Quote each relevant passage verbatim, including the message context 
   (who said it and when, if that information is present)
3. Do NOT paraphrase before quoting — always quote first, then explain
4. If the answer is not present in the conversation, say: 
   "This was not stated in the provided conversation."

My questions:
1. [QUESTION 1 — e.g., "What was said about the Q3 deadline?"]
2. [QUESTION 2]
3. [QUESTION 3]

Treat this as a retrieval task, not a generative one. 
Your only job is to find and quote what was actually said.
Never infer or add context from outside the conversation.

You are a decision-tracking assistant. I am providing a long conversation, 
document thread, or project history below.

[PASTE FULL CONTENT HERE]

Extract and organize every decision recorded in this material:

1. Decision Log
   For each decision:
   - Quote the decision statement verbatim (or as close as the text permits)
   - Note who made or ratified the decision (quote the attribution if present)
   - Note the date or conversation position (if timestamps or ordering are visible)
   - Note any stated conditions or expiry ("this applies until X" / "we'll revisit if Y")
   - Note if the decision was subsequently reversed — quote the reversal passage

2. Decision Map by Topic
   Group decisions by theme or project area.
   Within each group, order chronologically.

3. Conflicting Decisions
   Identify any cases where two decisions on the same topic contradict each other.
   Quote both. Note which came later.

4. Undocumented Decisions
   List any places where action was taken (tasks assigned, plans moved forward) 
   without a clearly stated decision. Quote the passage and note what decision seems implied.

Do not record informal suggestions or hypotheticals as decisions.
Only record what was committed to. If in doubt, quote and label it "Possible Decision — Ambiguous."

You are a conversation retrieval assistant. 
I am providing a complete conversation history below — this may be hundreds or thousands of messages.

[PASTE FULL CONVERSATION HISTORY HERE]

I need you to find everything I (the user) said about [TOPIC — e.g., "the pricing model" / 
"my preferred tech stack" / "the deadline for the v2 launch"].

For every message I sent that mentions or relates to [TOPIC]:
1. Quote my message verbatim
2. Note its position in the conversation 
   (e.g., "message 47 of 312" or the timestamp if visible)
3. Note any response it received — quote the response verbatim

Then produce:
- A chronological timeline of my stated views on [TOPIC]
- Where my position changed, quote both the earlier and later statement
- Where I asked a question about [TOPIC] that was not answered, flag it

This is a retrieval task only. Do not add interpretation or outside context.
If I never mentioned [TOPIC], say so explicitly — do not fabricate references.

You are a project communications specialist. 
I am providing the complete history of [PROJECT NAME] — 
all conversation threads, meeting notes, decisions, and relevant documents.
Each source is labeled === SOURCE: [Type — Date] ===

[PASTE ALL PROJECT HISTORY HERE]

A new stakeholder is joining [PROJECT NAME]. 
They have no prior context. Create a comprehensive onboarding brief.

1. Project Overview (150 words max)
   What is this project, what problem does it solve, and what is the current status?
   Source every claim with a quoted passage and its source label.

2. Key Decisions Made So Far
   List the 10 most important decisions with: the decision (verbatim quote), 
   the date/source, and a one-sentence explanation of why it matters.

3. Current State
   What is the team working on right now? What is the immediate next milestone?
   Quote the most recent source material that confirms this.

4. Open Issues and Blockers
   What is unresolved or at risk? Quote the source for each item.

5. Key People and Roles
   Who is involved and what is each person responsible for?
   Base this only on roles mentioned in the provided materials.

6. What the New Stakeholder Should Know That Isn't Written Down
   List anything that is implied by the documents but never explicitly stated — 
   label these as "Inferred from context" and quote the passages that suggest them.

Keep the brief grounded in the provided material. 
For every factual claim, cite the source. 
Where the record is incomplete, say so rather than filling gaps.