--- name: Research Gap Finder category: research works_with: [claude, chatgpt, gemini] difficulty: plug-and-play version: 1.0 source: adapted (user-submitted collection, rewritten) tested: no --- # Research Gap Finder ## What it does Maps the research landscape for a topic and pinpoints where the gaps, weak methods, and outdated assumptions are — so you know where the unexplored, high-value questions sit. ## When to use it - Starting a literature review and deciding where to focus - Looking for a thesis, article, or content angle nobody has covered - Checking whether "everyone knows X" in a field is actually supported ## The skill ``` You are a research methodologist. Topic: [RESEARCH TOPIC] Work through this in order: 1. LANDSCAPE — Summarize the current state of research in 5-8 bullet points: dominant theories, key findings, main camps. 2. GAPS — Identify the 3 biggest literature gaps. For each: what's missing, why it matters, and why it's likely been overlooked. 3. WEAK METHODS — List the most common methodological weaknesses in this field (sample sizes, measurement, replication, conflicts of interest). 4. OUTDATED ASSUMPTIONS — Name beliefs the field still repeats that newer evidence questions. 5. DIRECTIONS — Rank the 3 highest-impact research directions by (a) importance and (b) feasibility. Explain the ranking. Rules: cite specific studies or reviews where you can; clearly flag anything you're uncertain about instead of presenting it as fact. ``` ## Example output [TO FILL AFTER TESTING] ## Tweaks - Add "focus on the last 5 years only" for fast-moving fields - Add your intended use ("I'm writing a YouTube script on this") to bias the directions toward your goal - Run with web search on for citations you can verify --- --- name: Concept Matrix Builder category: research works_with: [claude, chatgpt, gemini] difficulty: plug-and-play version: 1.0 source: adapted (user-submitted collection, rewritten) tested: no --- # Concept Matrix Builder ## What it does Turns a pile of source material into one structured comparison table — definitions, methods, metrics, and patterns — with the repetitive filler stripped out. ## When to use it - You've collected 5+ articles/papers/transcripts and need them condensed - Preparing to write and you want your raw material organized first - Comparing frameworks or tools that all describe themselves differently ## The skill ``` You are an information architect. Source material: [PASTE OR ATTACH] Build: 1. A comparison table. Rows = each source. Columns: Core claim | Method used | Key metric(s) | Strongest evidence | Weakness 2. DEFINITIONS — The 5-10 terms these sources use differently or loosely, with each source's working definition. 3. PATTERNS — 3-5 patterns that appear across multiple sources that no single source states outright. 4. DISCARD PILE — One line listing what you removed as repetitive or low-value, so I know nothing important was silently dropped. Rules: table cells max 15 words. If sources conflict, show the conflict in the table rather than averaging it away. ``` ## Example output [TO FILL AFTER TESTING] ## Tweaks - Add a "Relevance to [MY PROJECT]" column to make it decision-ready - For video research, add: "mark the 3 most surprising cells — those are hooks" --- --- name: Batch Paper Review SOP category: research works_with: [claude, chatgpt, gemini] difficulty: needs-setup version: 1.0 source: adapted (user-submitted collection, rewritten) tested: no --- # Batch Paper Review SOP ## What it does A repeatable procedure for reviewing up to 10 papers in one pass, producing an identical structure for every paper so you can scan and compare at a glance. ## When to use it - Literature review sessions where volume matters more than depth - Weekly "keep up with the field" reading - Building a research base for long-form content ## The skill ``` You are a research analyst. I'm giving you [N] papers: [ATTACH/PASTE] For EACH paper, use exactly this structure: PAPER [#]: [Short title] DOMAIN: [field] | DATASET: [what was studied, n=] FINDING: One sentence, the single most important result. METHOD: One sentence. STRENGTH: strong / moderate / weak — with a 5-word reason. STANDOUT: The one detail worth remembering. Then after all papers: CROSS-CUTTING TRENDS: 3 bullets on what the batch collectively shows. CONFLICTS: Where papers disagree. TOP 3: The papers most worth reading in full, and why in one line each. Rules: identical structure every paper, no exceptions. If a paper doesn't state something (like n=), write "not reported" — never guess. ``` ## Example output [TO FILL AFTER TESTING] ## Tweaks - Reduce to 3 fields per paper for 20+ paper batches - Add "flag any paper funded by industry with a conflict of interest" - Save the output — batches stack into a searchable personal database over time --- --- name: Complex Topic Explainer category: content works_with: [claude, chatgpt, gemini] difficulty: plug-and-play version: 1.0 source: adapted (user-submitted collection, rewritten) tested: no --- # Complex Topic Explainer ## What it does Explains a technical or abstract topic in plain spoken language with analogies, so you actually understand it — or can teach it in your own content. ## When to use it - Researching a topic you need to explain on camera or in writing - A paper/doc is full of jargon and you need the actual idea - Checking whether YOU understand something well enough to teach it ## The skill ``` You are an educator who explains things the way a smart friend would over coffee. Topic: [COMPLEX TOPIC] 1. THE ONE-LINER — The whole idea in one sentence a 15-year-old gets. 2. THE ANALOGY — One everyday analogy that carries the core mechanism. Then one sentence on where the analogy breaks down. 3. THE WALKTHROUGH — Explain it properly in under 300 words. Every technical term gets defined in-line the first time it appears. 4. WHY IT MATTERS — 2-3 practical implications for a normal person. 5. THE TRAP — The most common way people misunderstand this topic. Rules: conversational language, no bullet spam, no "essentially" or "simply put". If a formula is involved, explain what each part means in words before showing it. ``` ## Example output [TO FILL AFTER TESTING] ## Tweaks - Add "my audience is [X]" to tune the analogies - Add "give me 3 analogy options" if the first one feels flat - For scripts: ask for THE TRAP first — misconceptions make strong hooks --- --- name: Conflicting Evidence Analyzer category: research works_with: [claude, chatgpt, gemini] difficulty: plug-and-play version: 1.0 source: adapted (user-submitted collection, rewritten) tested: no --- # Conflicting Evidence Analyzer ## What it does Takes multiple studies that disagree and works out WHY they disagree — then tells you what conclusion the evidence actually supports. ## When to use it - "Study says X" vs "study says opposite of X" situations - Health/fitness/nutrition topics where headlines contradict - Before publicly claiming anything research-based ## The skill ``` You are a research analyst. Studies/claims to analyze: [PASTE/ATTACH] 1. THE DISAGREEMENT — State precisely what the sources disagree on (often narrower than it first appears). 2. WHY THEY DIFFER — Check each explanation: different populations, doses/durations, measurements, definitions, funding, or era. Identify which applies here. 3. EVIDENCE WEIGHT — Rank the sources by strength: study design, sample size, replication, independence. Strongest first. 4. ANOMALIES — Anything that doesn't fit the pattern and deserves suspicion. 5. VERDICT — The most defensible conclusion, stated with calibrated confidence: "strong evidence that…", "weak evidence that…", or "genuinely unresolved". Never force a winner if there isn't one. Rules: "unresolved" is an acceptable answer. Flag any point where you're inferring rather than reading directly from the sources. ``` ## Example output [TO FILL AFTER TESTING] ## Tweaks - Add "explain the verdict in 2 sentences I could say on camera" - Combine with Source Credibility Auditor when sources look sketchy --- --- name: Steelman & Verdict category: research works_with: [claude, chatgpt, gemini] difficulty: plug-and-play version: 1.0 source: adapted (user-submitted collection, rewritten) tested: no --- # Steelman & Verdict ## What it does Builds the strongest possible case AGAINST a claim you believe, then weighs both sides and gives the most defensible position — your protection against confirmation bias. ## When to use it - Before building content or a business decision on a belief - Testing your own conclusions after a research session - Preparing for objections your audience will raise ## The skill ``` You are an academic reviewer with no loyalty to my position. Claim: [CLAIM] STEP 1 — STEELMAN: Build the strongest evidence-based counterargument against this claim. Argue it like you believe it. Use the best available evidence, not strawman versions. STEP 2 — WEIGH: - Strengths and weaknesses of each side - Hidden assumptions each side relies on - What evidence is missing that would settle it STEP 3 — VERDICT: The most defensible position, which may be the claim, the counter, or a conditional middle ("true for X, not for Y"). End with: "What would change this verdict: ..." — the specific evidence that would flip it. Rules: do not soften the counterargument to spare my feelings. If my claim loses, say so plainly. ``` ## Example output [TO FILL AFTER TESTING] ## Tweaks - Run on your video thesis before scripting — objections become content - Add "limit to 400 words" for a fast gut-check version --- --- name: Source Credibility Auditor category: research works_with: [claude, chatgpt, gemini] difficulty: plug-and-play version: 1.0 source: adapted (user-submitted collection, rewritten) tested: no --- # Source Credibility Auditor ## What it does Runs your sources through a structured credibility check and rates each one, so weak or biased sources don't sneak into your work. ## When to use it - Before citing anything publicly - When a source confirms what you WANT to be true (highest-risk moment) - Auditing research an AI did for you ## The skill ``` You are a research auditor. Sources to evaluate: [LIST/PASTE/ATTACH] For EACH source, score 1-5 on: - AUTHORITY: author/institution track record in THIS field - METHOD: how the claims were produced (study > expert view > opinion) - INDEPENDENCE: funding, incentives, conflicts of interest - CORROBORATION: do independent sources reach the same conclusion? - CURRENCY: is it current for a field that moves at this speed? Output a table: Source | A | M | I | C | C | Overall | One-line reason. Then: - RED FLAGS: anything that pattern-matches to marketing dressed as research, cherry-picking, or citation laundering (sources citing each other in a circle). - MY BIAS CHECK: which of these sources I'm most likely trusting because it agrees with me, based on what I've said. - KEEP/DROP: which sources are safe to build on. Rules: a famous name is not authority outside their field. When you can't verify something about a source, score it down, don't guess up. ``` ## Example output [TO FILL AFTER TESTING] ## Tweaks - Add your claim/thesis so the bias check has something to work with - For fast checks: "audit only the 3 sources I rely on most" --- --- name: Research Session Debrief category: research works_with: [claude, chatgpt, gemini] difficulty: plug-and-play version: 1.0 source: adapted (user-submitted collection, rewritten) tested: no --- # Research Session Debrief ## What it does A post-session audit of your research workflow — what you found, what you missed, and what to change — so each session gets faster and sharper than the last. ## When to use it - End of a timeboxed research sprint (20-60 min) - Your research sessions feel busy but produce little - Tuning your prompts and process over time ## The skill ``` You are a workflow auditor. Here's my research session: - Time spent: [X minutes] - Goal: [WHAT I WAS TRYING TO ANSWER] - Sources reviewed: [LIST OR COUNT] - What I concluded: [PASTE NOTES/OUTPUT] Audit: 1. GOAL HIT? Did the output actually answer the goal — fully, partially, or did I drift? 2. YIELD: insights per source. Which sources produced nothing and what did they have in common (so I stop picking them)? 3. ERRORS: anything in my conclusions that's unsupported, overstated, or contradicted by my own sources. 4. BOTTLENECK: the single step that consumed the most time for the least value. 5. NEXT SESSION: 3 specific changes — a better prompt, a source type to skip, a structure to reuse. Concrete, not "be more focused". Rules: be blunt. A session that produced a wrong conclusion quickly is worse than a slow one that produced a right one. ``` ## Example output [TO FILL AFTER TESTING] ## Tweaks - Run weekly on your best and worst session and compare - Feed the "next session" changes into your notes so they compound