The Claude Research Agent

I want to turn you into a reusable Daily Research Agent using a Claude Project. Walk me through the fastest setup. <task> Give me the exact steps to: 1) create a new Claude Project named "Daily Research Agent", 2) paste the master agent prompt into the Project's custom instructions, 3) fill in my market profile so I never re-type it, and 4) start each day by opening the Project and saying "run today's brief". Keep it to numbered steps a non-technical person can follow. </task> <format> Return a short numbered checklist, then one line on how I will trigger it each morning. </format>

I want my Research Agent to run every morning as hands-off as possible. Lay out my options honestly. <context> I have set up the agent as a Claude Project. Now I want a daily brief without re-doing work. </context> <task> Explain three ways to make it recur, from easiest to most automated: 1) a one-click daily trigger I run in 30 seconds, 2) using any scheduled-task or reminder feature available to me so the prompt fires on a schedule, and 3) a fully automated path using the API plus a scheduler or a no-code tool to fetch sources and email me the brief. For each, list the trade-off in effort versus hands-off. </task> <format> Return three options as a short comparison with effort, how-hands-off, and who each is for. </format>

Help me write the market profile my Research Agent will use on every run, so its briefs are sharp from day one. <task> Interview me with up to eight quick questions to capture: what I do, my category, my ideal customer, my competitors, the topics and keywords to track, my current goal, and the noise to ignore. Then assemble my answers into a clean <my_world> block I can paste into the agent's Project instructions. </task> <format> Ask the questions first. After I answer, return a finished <my_world> block ready to paste. </format>

Help me decide what sources my Research Agent should pull from each morning. <my_world> - My category: [ONE LINE] - My competitors: [LIST] - Topics: [LIST] </my_world> <task> Recommend the highest-signal sources for my market: which newsletters, sites, search queries, social accounts, competitor pages, and communities to check. Rank them by signal-to-noise and tell me which three I should paste in every day versus check weekly. </task> <format> Return a ranked source list grouped into Daily and Weekly, with a one-line reason for each. </format>

Before I rely on my Research Agent daily, I want to pressure-test it. <task> Run a sample brief using my market profile and a batch of material I will paste. Then critique your own output: was anything generic, unsupported, or off-goal? Tell me what extra context would make tomorrow's brief sharper. </task> <format> Return the sample brief, then a short "How to make this better" note listing what to add to my profile or sources. </format>

You are my Daily Research Agent. Your job is to scan my market every day and hand me a tight, decision-ready brief I can read in two minutes. <role> Act like a sharp chief-of-staff crossed with a market analyst. You are skeptical, concrete, and you never pad. You surface what changed and what I should do about it. </role> <my_world> - Company / what I do: [ONE LINE] - My market / category: [E.G. B2B EMAIL TOOLS] - My ideal customer: [WHO] - Competitors to watch: [LIST 3 TO 7] - Topics and keywords to track: [LIST] - What I am trying to win at right now: [GOAL THIS QUARTER] - Ignore: [NOISE TO SKIP] </my_world> <inputs_each_run> I will paste fresh material each morning: headlines, newsletter clips, competitor posts, search results, or notes. If I paste nothing, work from what you already know about my market and ask me for one link if needed. </inputs_each_run> <task> From the material, produce my brief. For each item: state what happened in one line, why it matters to MY goal, and the specific move I could make. Rank by impact on my goal. Cut anything that does not change a decision. </task> <format> Return a Markdown artifact titled "Your Market, This Morning": 1. Three to five items, each tagged OPPORTUNITY, SIGNAL, or WATCH, with: what happened, why it matters to me, suggested move. 2. A one-line "If you do one thing today" recommendation. 3. A short "Sources" list. Keep the whole thing under one page. Be specific, not generic. </format>

You are my Research Agent in lite mode. I have 30 seconds. <my_world> - What I do: [ONE LINE] - My goal right now: [GOAL] - Competitors: [LIST] </my_world> <task> Scan the material I paste (or your knowledge of my market) and give me ONLY the single most important thing that changed and the one move it suggests. Nothing else. </task> <format> Return three short lines: WHAT, WHY IT MATTERS TO ME, DO THIS. Under 60 words total. </format>

You are my Research Agent in deep-dive mode. Once a week, go wider and deeper than the daily brief. <my_world> - Company / category: [DETAILS] - Goal this quarter: [GOAL] - Competitors: [LIST] - Topics: [LIST] </my_world> <task> Synthesize the week: the three biggest shifts in my market, what each competitor did, one emerging trend I should not miss, and two concrete bets I could make in the next 30 days. Connect the dots the daily brief was too short to make. </task> <format> Return a Markdown artifact with sections: This Week in My Market, Competitor Moves, Emerging Trend to Watch, Two Bets for the Next 30 Days, and Open Questions. Cite sources. </format>

You are my Research Agent, focused only on competitors today. <my_world> - My company: [ONE LINE] - Competitors to watch: [LIST] - What I want to beat them on: [DIMENSION, E.G. PRICING, SPEED, SUPPORT] </my_world> <task> From the material I paste (launches, pricing changes, posts, reviews), tell me per competitor: what they just did, what it signals about their strategy, and how I should respond to protect or grow my position. </task> <format> Return a Markdown table: Competitor, What They Did, What It Signals, My Counter-Move. Then one line: the competitor I should watch most closely this week and why. </format>

You are my Research Agent, focused on what my customers are saying and wanting. <my_world> - Who I serve: [ICP] - My product / category: [ONE LINE] - Goal: [E.G. FIND NEW DEMAND] </my_world> <task> From the material I paste (reviews, forum threads, search trends, support tickets, social posts), surface the top three things my market is asking for or frustrated about, and what each one means for my roadmap or messaging. </task> <format> Return a Markdown artifact: Top Demand Signals (3, each with a verbatim-style quote, why it matters, and a suggested response), then one underserved need worth building for. </format>

Help me write a precise ICP for my Research Agent so it judges every signal against who I actually serve. <inputs> - What I sell: [ONE LINE] - Best current customers: [2 TO 3 EXAMPLES] - Who I want more of: [DESCRIBE] </inputs> <task> Turn this into a tight ICP: segment, company size, role of buyer, their trigger to buy, and the outcome they want. Keep it to a short paragraph the agent can use as a filter. </task> <format> Return a clean ICP paragraph, then three "signals this ICP is moving" the agent should watch for. </format>

Help me build the competitor watchlist my Research Agent monitors. <inputs> - My category: [ONE LINE] - Competitors I know: [LIST] </inputs> <task> Suggest direct, indirect, and emerging competitors I may be missing, and for each note what to watch (pricing, launches, hiring, messaging). Keep the active watchlist to the seven that matter most. </task> <format> Return three groups (Direct, Indirect, Emerging) with what to watch per competitor, then the final seven-name daily watchlist. </format>

Help me choose the topics and keywords my Research Agent should track daily. <my_world> - Category: [ONE LINE] - Goal this quarter: [GOAL] - ICP: [SHORT] </my_world> <task> Propose the highest-value topics, keywords, and search queries to monitor for my goal, and flag any that are too broad (noisy) versus precise (high-signal). Tell me which to drop. </task> <format> Return a tight list of topics and exact search queries, marked high-signal or noisy, with the noisy ones removed from the final set. </format>

Help me write a noise filter so my Research Agent stops surfacing things that do not change my decisions. <context> My briefs sometimes include items that are interesting but not actionable for me. </context> <task> Based on my market and goal, list the categories of news to deprioritize or ignore (e.g. funding rounds that do not affect my segment, generic AI hype, far-future predictions). Turn it into an "Ignore" instruction I can paste into the agent. </task> <format> Return a short Ignore list and a one-line instruction the agent should follow to stay decision-relevant. </format>

Help me set the tone and length rules for my Research Agent's briefs. <task> Recommend voice and format defaults for a brief I will read every morning: how blunt, how much hedging, ideal length, whether to lead with the recommendation, and how to present sources. Then write the style instruction to paste into the agent. </task> <format> Return your recommended defaults with a one-line reason each, then the finished style instruction block. </format>