Research Assistant Agent Loop With Source Tracking
Define a plan, search, read, note and verify loop that never invents a source.
Added October 15, 2026. Written with AI assistance and checked for clarity and placeholders; not reviewed by a lawyer, accountant, clinician or other professional. How these prompts are made
The Open buttons send the prompt text as currently shown below to that site only when you click. Fill in the placeholders first.
Design a research agent loop for [RESEARCH_TOPIC] for a reader who is [AUDIENCE]. Tools available: [TOOLS_AVAILABLE]. Time and length limits: [LIMITS]. Quality bar: [QUALITY_BAR]. Specify the loop as numbered steps with a stopping rule: 1) restate the question and list sub-questions; 2) choose search queries and say why; 3) for each source record the title, author or organisation, date, URL or location, what it claims, and what kind of source it is (primary, review, news, opinion); 4) extract claims into a ledger with a direct quote under 20 words and the page or section; 5) mark each claim confirmed by one source, two independent sources or contradicted; 6) stop when each sub-question has two independent sources or a clear note that none exist; 7) write the summary from the ledger only. Provide the ledger table layout, the rule that any claim without a ledger entry is cut, a template for 'what I could not find', and how to handle paywalled or unreadable sources honestly. Also write the agent's final answer format with citations by ledger number. Do not let it cite from memory.
Fill in before use
Optional: type your details and the prompt above updates. Nothing is saved or sent.
Replace each bracketed placeholder with your own details, then paste into Any AI or another assistant.