{"id":"7c54a423-5940-486f-a838-3eeb00c9b304","shortId":"8MLxDQ","kind":"skill","title":"research-synthesizer","tagline":"Gathers information from multiple sources in parallel, compares what they say, flags where they agree or conflict, and delivers a clean synthesis with a confidence rating for each finding. Useful for making evidence-backed decisions, when evaluating competing claims, during due d","description":"# Research Synthesizer\n\n## Role\n\nYou are a research director coordinating a team of parallel workers. You\ndispatch each worker to a different source, collect what they find, surface\nwhere the evidence converges and where it conflicts, and present the\nsynthesized result in a format the user can act on immediately.\n\n## Why This Skill Exists\n\nSingle-source research is fast but fragile. When every finding comes from\none article or one search result, you cannot tell what's broadly true vs\nwhat's one author's opinion. Multi-source synthesis separates signal from\nnoise — and it shows the user exactly how confident to be in each finding.\n\nManual multi-source research takes hours. Cowork with parallel workers does\nit in minutes.\n\n## Instructions\n\n### Step 1 — Clarify the research question and scope\n\nBefore dispatching any workers, confirm:\n\n1. What is the specific question or topic? (The tighter the question, the\n   more useful the synthesis.)\n2. What kind of sources should be prioritized — recent news, academic\n   research, industry reports, product documentation, public company data?\n3. How many sources? Recommend 4–6 for most topics. More than 8 adds noise.\n4. What is the deliverable — a one-page brief, a comparison table, a\n   detailed report?\n5. Is there a time horizon? (\"Last 12 months\" vs \"all time\" changes which\n   sources matter.)\n\nWrite the confirmed research brief as a single paragraph before proceeding.\n\n### Step 2 — Identify the source types\n\nChoose 4–6 source types appropriate to the question. Common combinations:\n\n- For market research: industry reports, competitor websites, news articles,\n  analyst commentary, customer reviews\n- For decision support: expert opinion pieces, case studies, official\n  documentation, statistical data\n- For competitive intelligence: company websites, job postings, press\n  releases, product reviews, social commentary\n\nName each source type explicitly. Workers perform better with a specific\nassignment than a vague one.\n\n### Step 3 — Dispatch parallel workers\n\nWrite one research prompt per source type, then combine them into a single\nparallel dispatch instruction:\n\n> \"Use parallel workers to research [topic]. Assign one worker per source\n> type below. Each worker should gather the 3–5 most relevant pieces of\n> information from their assigned source type and save findings to a file\n> named research-[source-type].md. Workers should not share findings with\n> each other yet.\n>\n> Source assignments:\n> - Worker 1: [source type 1]\n> - Worker 2: [source type 2]\n> - Worker 3: [source type 3]\n> - Worker 4: [source type 4]\"\n\n### Step 4 — Collect and compare the raw findings\n\nOnce all workers have saved their files, read all research files and perform\na cross-source comparison. For each major theme or claim, note:\n\n- Which sources agree\n- Which sources contradict each other\n- Which sources provide data vs opinion\n- What is missing from all sources\n\nSave this comparison to `research-comparison.md`.\n\n### Step 5 — Synthesize into findings with confidence ratings\n\nWrite the final synthesis. For each key finding, assign a confidence rating:\n\n- **High confidence:** Three or more independent sources agree, at least one\n  is data-backed.\n- **Medium confidence:** Two sources agree, or one strong source with no\n  contradictions found.\n- **Low confidence:** Only one source, or sources conflict without resolution.\n\nPresent each finding as: [Finding statement] — [Confidence: High/Medium/Low]\nfollowed by one sentence explaining why.\n\n### Step 6 — Deliver the synthesis report\n\nStructure the final report as:\n\n1. **Research question** (one sentence)\n2. **Key findings** (5–8 bullet points with confidence ratings)\n3. **Points of disagreement** (where sources conflict and why it matters)\n4. **Gaps** (what was not found and where to look next)\n5. **Recommended action** (what the user should do with this information)\n\nSave to `research-synthesis-[topic]-[date].md` and present the key findings\ndirectly in the conversation.\n\n### Step 7 — Offer a follow-up drill-down\n\nAsk: \"Is there any finding you want to go deeper on? I can dispatch a\nfocused worker on that specific question.\"\n\nThis turns a broad synthesis into a targeted investigation without starting over.\n\n## Compounds With\n\n- **parallel-power** — the parallel dispatch in Step 3 is the core engine\n- **connected tools** — web search, document access, and database connections\n  all expand the source range available to workers\n- **output formatting** — the final synthesis can be reformatted into a\n  slide deck, email brief, or client report\n\n## Quality Checks\n\nBefore finishing, verify:\n\n- [ ] Research question is specific enough to yield actionable findings\n- [ ] At least 4 distinct source types were assigned to separate workers\n- [ ] All worker output files were saved before synthesis began\n- [ ] Every key finding has a confidence rating with a one-sentence rationale\n- [ ] Points of disagreement between sources are explicitly called out\n- [ ] Gaps (what was not found) are listed — not hidden\n- [ ] The final report is saved to a named file, not just shown in chat\n- [ ] A follow-up drill-down offer was made\n\n## Output Format\n\n```\n# Research Synthesis: [Topic]\nResearch question: [One-sentence statement of what was investigated]\nDate: [Date] | Sources queried: [number] | Scope: [time horizon]\n\n---\n\n## Key Findings\n\n1. [Finding statement] — **Confidence: High**\n   *[One sentence explaining why: which sources agreed, whether data-backed]*\n\n2. [Finding statement] — **Confidence: Medium**\n   *[One sentence explaining why: sources that agreed, any caveats]*\n\n3. [Finding statement] — **Confidence: Low**\n   *[One sentence explaining why: single source, conflict, or limited data]*\n\n[Continue for 5–8 total findings]\n\n---\n\n## Points of Disagreement\n\n**[Topic of disagreement]:**\n- Source A says: [position]\n- Source B says: [position]\n- Why it matters: [implication for the user's question]\n\n[Repeat for each material conflict]\n\n---\n\n## Gaps\n\nWhat was not found and where to look next:\n- [Gap 1] — Suggested next source: [where to look]\n- [Gap 2] — Suggested next source: [where to look]\n\n---\n\n## Recommended Action\n\n[1–3 sentences. What the user should do, decide, or investigate further based on the synthesis.\nGrounded in the high-confidence findings.]\n\n---\n\n*Sources queried: [list source types used]*\n*Files saved: research-[source]-[date].md (per worker), research-comparison.md, this file*\n```\n\n## Examples\n\n**Example 1 — Market research:**\nUser types: \"Research synthesizer — is there demand for async video tools in mid-market B2B?\"\nResult: Cowork dispatches four workers (industry reports, competitor analysis, customer reviews, analyst\ncommentary), collects findings, compares where sources agree and conflict, and delivers a synthesis with\nconfidence-rated findings — saving the full report as `research-synthesis-async-video-b2b-[date].md`.\n\n**Example 2 — Competitive intelligence:**\nUser types: \"Deep dive into how Notion positions itself vs. Confluence for enterprise teams\"\nResult: Workers cover company websites, job postings, press releases, product reviews, and social\ncommentary. The synthesis surfaces where positioning claims align with user perception and where they\ndiverge — flagging gaps in the competitive story with a confidence rating for each finding.\n\n**Example 3 — Decision support:**\nUser types: \"What does the evidence say about four-day work weeks and productivity?\"\nResult: Cowork prioritizes academic research, expert commentary, and case studies from companies that\nhave tried it. The synthesis separates high-confidence findings (backed by multiple studies) from low-\nconfidence claims (single source or methodologically weak), giving the user a clear picture of what the\nevidence actually supports.\n\n## Troubleshooting\n\n**Issue: \"Confidence ratings all came back Low — nothing feels settled\"**\nThis is accurate data, not a failure. It means the topic is genuinely contested or under-researched. The\nsynthesis is still valuable: it maps the landscape of disagreement and shows where more authoritative\nsources are needed. Use the Gaps section and drill-down offer to identify the next research step.\n\n**Issue: \"Workers found conflicting information and I don't know which to trust\"**\nThe Points of Disagreement section exists for exactly this. Look at the source type behind each claim:\ndata-backed findings from primary research outweigh opinion pieces. If two credible sources conflict,\nthe finding gets a Low or Medium confidence rating and a note — so you can see the disagreement rather\nthan have it hidden in a blended average.\n\n**Issue: \"The synthesis is too broad — I need more depth on one finding\"**\nUse the drill-down offer at Step 7. Say: \"Go deeper on finding 3.\" The skill dispatches a focused worker\non that specific question, running a targeted investigation without restarting the full synthesis.\n\n## Related Skills\n\nSee also: **content-engine** — uses research synthesis as the foundation for blog posts and social content.\nRelated: **weekly-business-pulse** — applies a similar multi-source aggregation pattern to your own tools.\nRelated: **meeting-machine** — for bringing synthesized research into a pre-meeting prep package.","tags":["research","synthesizer","cowork","skills","library","eaiconsulting","agent-skills","ai-automation","ai-skills","anthropic-claude","claude","claude-code-plugin"],"capabilities":["skill","source-eaiconsulting","skill-research-synthesizer","topic-agent-skills","topic-ai-automation","topic-ai-skills","topic-anthropic-claude","topic-claude","topic-claude-code-plugin","topic-claude-code-skills","topic-claude-cowork","topic-claude-desktop","topic-claude-skills","topic-cowork","topic-knowledge-work"],"categories":["cowork-skills-library"],"synonyms":[],"warnings":[],"endpointUrl":"https://skills.sh/EAIconsulting/cowork-skills-library/research-synthesizer","protocol":"skill","transport":"skills-sh","auth":{"type":"none","details":{"cli":"npx skills add EAIconsulting/cowork-skills-library","source_repo":"https://github.com/EAIconsulting/cowork-skills-library","install_from":"skills.sh"}},"qualityScore":"0.453","qualityRationale":"deterministic score 0.45 from registry signals: · indexed on github topic:agent-skills · 7 github stars · SKILL.md body (9,228 chars)","verified":false,"liveness":"unknown","lastLivenessCheck":null,"agentReviews":{"count":0,"score_avg":null,"cost_usd_avg":null,"success_rate":null,"latency_p50_ms":null,"narrative_summary":null,"summary_updated_at":null},"enrichmentModel":"deterministic:skill-github:v1","enrichmentVersion":1,"enrichedAt":"2026-05-18T19:13:34.542Z","embedding":null,"createdAt":"2026-05-18T13:21:05.771Z","updatedAt":"2026-05-18T19:13:34.542Z","lastSeenAt":"2026-05-18T19:13:34.542Z","tsv":"'1':171,183,423,426,583,854,944,961,1003 '12':257 '2':200,278,428,431,588,870,952,1066 '3':219,349,387,433,436,598,701,884,962,1125,1353 '4':224,234,284,438,441,443,609,756 '5':250,388,501,591,620,901 '6':225,285,573 '7':649,1347 '8':231,592,902 'academ':210,1146 'access':711 'accur':1204 'act':93 'action':622,752,960 'actual':1189 'add':232 'aggreg':1403 'agre':18,477,527,539,865,881,1040 'align':1103 'also':1376 'analysi':1030 'analyst':303,1033 'appli':1397 'appropri':288 'articl':114,302 'ask':658 'assign':343,375,396,421,516,761 'async':1014,1060 'author':130 'authorit':1235 'avail':720 'averag':1325 'b':916 'b2b':1021,1062 'back':38,534,869,1166,1197,1286 'base':973 'began':773 'behind':1281 'better':339 'blend':1324 'blog':1387 'brief':243,270,736 'bring':1414 'broad':124,682,1331 'bullet':593 'busi':1395 'call':794 'came':1196 'cannot':120 'case':313,1151 'caveat':883 'chang':262 'chat':818 'check':741 'choos':283 'claim':43,473,1102,1173,1283 'clarifi':172 'clean':24 'clear':1183 'client':738 'collect':69,444,1035 'combin':293,361 'come':111 'commentari':304,331,1034,1096,1149 'common':292 'compani':217,322,1086,1154 'compar':11,446,1037 'comparison':245,467,497 'compet':42 'competit':320,1067,1115 'competitor':299,1029 'compound':691 'confid':28,148,506,518,521,536,549,564,596,779,857,873,887,982,1049,1119,1164,1172,1193,1306 'confidence-r':1048 'confirm':182,268 'conflict':20,81,555,604,895,932,1042,1257,1298 'confluenc':1079 'connect':706,714 'content':1378,1391 'content-engin':1377 'contest':1215 'continu':899 'contradict':480,546 'converg':77 'convers':647 'coordin':55 'core':704 'cover':1085 'cowork':161,1023,1144 'credibl':1296 'cross':465 'cross-sourc':464 'custom':305,1031 'd':46 'data':218,318,486,533,868,898,1205,1285 'data-back':532,867,1284 'databas':713 'date':637,844,845,994,1063 'day':1138 'decid':969 'decis':39,308,1126 'deck':734 'deep':1071 'deeper':667,1350 'deliv':22,574,1044 'deliver':238 'demand':1012 'depth':1335 'detail':248 'differ':67 'direct':644 'director':54 'disagr':601,789,907,910,1230,1270,1316 'dispatch':62,179,350,367,671,698,1024,1356 'distinct':757 'dive':1072 'diverg':1110 'document':215,316,710 'drill':656,824,1245,1342 'drill-down':655,823,1244,1341 'due':45 'email':735 'engin':705,1379 'enough':749 'enterpris':1081 'evalu':41 'everi':109,774 'evid':37,76,1133,1188 'evidence-back':36 'exact':146,1274 'exampl':1001,1002,1065,1124 'exist':99,1272 'expand':716 'expert':310,1148 'explain':570,861,877,891 'explicit':336,793 'failur':1208 'fast':105 'feel':1200 'file':404,456,460,768,813,990,1000 'final':510,580,726,806 'find':32,72,110,153,401,415,449,504,515,560,562,590,643,662,753,776,853,855,871,885,904,983,1036,1051,1123,1165,1287,1300,1338,1352 'finish':743 'flag':15,1111 'focus':673,1358 'follow':566,653,821 'follow-up':652,820 'format':89,724,830 'found':547,614,800,937,1256 'foundat':1385 'four':1025,1137 'four-day':1136 'fragil':107 'full':1054,1371 'gap':610,796,933,943,951,1112,1241 'gather':4,385 'genuin':1214 'get':1301 'give':1179 'go':666,1349 'ground':977 'hidden':804,1321 'high':520,858,981,1163 'high-confid':980,1162 'high/medium/low':565 'horizon':255,851 'hour':160 'identifi':279,1249 'immedi':95 'implic':922 'independ':525 'industri':212,297,1027 'inform':5,393,630,1258 'instruct':169,368 'intellig':321,1068 'investig':687,843,971,1367 'issu':1192,1254,1326 'job':324,1088 'key':514,589,642,775,852 'kind':202 'know':1263 'landscap':1228 'last':256 'least':529,755 'limit':897 'list':802,986 'look':618,941,950,958,1276 'low':548,888,1171,1198,1303 'machin':1412 'made':828 'major':470 'make':35 'mani':221 'manual':154 'map':1226 'market':295,1004,1020 'materi':931 'matter':265,608,921 'md':410,638,995,1064 'mean':1210 'medium':535,874,1305 'meet':1411,1421 'meeting-machin':1410 'methodolog':1177 'mid':1019 'mid-market':1018 'minut':168 'miss':491 'month':258 'multi':134,156,1401 'multi-sourc':133,155,1400 'multipl':7,1168 'name':332,405,812 'need':1238,1333 'news':209,301 'next':619,942,946,954,1251 'nois':140,233 'note':474,1310 'noth':1199 'notion':1075 'number':848 'offer':650,826,1247,1344 'offici':315 'one':113,116,129,241,347,354,376,530,541,551,568,586,784,837,859,875,889,1337 'one-pag':240 'one-sent':783,836 'opinion':132,311,488,1292 'output':723,767,829 'outweigh':1291 'packag':1423 'page':242 'paragraph':274 'parallel':10,59,163,351,366,370,694,697 'parallel-pow':693 'pattern':1404 'per':357,378,996 'percept':1106 'perform':338,462 'pictur':1184 'piec':312,391,1293 'point':594,599,787,905,1268 'posit':914,918,1076,1101 'post':325,1089,1388 'power':695 'pre':1420 'pre-meet':1419 'prep':1422 'present':83,558,640 'press':326,1090 'primari':1289 'priorit':207,1145 'proceed':276 'product':214,328,1092,1142 'prompt':356 'provid':485 'public':216 'puls':1396 'qualiti':740 'queri':847,985 'question':175,188,194,291,585,678,746,835,927,1363 'rang':719 'rate':29,507,519,597,780,1050,1120,1194,1307 'rather':1317 'rational':786 'raw':448 'read':457 'recent':208 'recommend':223,621,959 'reformat':730 'relat':1373,1392,1409 'releas':327,1091 'relev':390 'repeat':928 'report':213,249,298,577,581,739,807,1028,1055 'research':2,47,53,103,158,174,211,269,296,355,373,406,459,584,634,745,831,834,992,1005,1008,1058,1147,1219,1252,1290,1381,1416 'research-comparison.md':499,998 'research-synthes':1 'research-synthesi':633 'research-synthesis-async-video-b2b':1057 'resolut':557 'restart':1369 'result':86,118,1022,1083,1143 'review':306,329,1032,1093 'role':49 'run':1364 'save':400,454,495,631,770,809,991,1052 'say':14,913,917,1134,1348 'scope':177,849 'search':117,709 'section':1242,1271 'see':1314,1375 'sentenc':569,587,785,838,860,876,890,963 'separ':137,763,1161 'settl':1201 'share':414 'show':143,1232 'shown':816 'signal':138 'similar':1399 'singl':101,273,365,893,1174 'single-sourc':100 'skill':98,1355,1374 'skill-research-synthesizer' 'slide':733 'social':330,1095,1390 'sourc':8,68,102,135,157,204,222,264,281,286,334,358,379,397,408,420,424,429,434,439,466,476,479,484,494,526,538,543,552,554,603,718,758,791,846,864,879,894,911,915,947,955,984,987,993,1039,1175,1236,1279,1297,1402 'source-eaiconsulting' 'source-typ':407 'specif':187,342,677,748,1362 'start':689 'statement':563,839,856,872,886 'statist':317 'step':170,277,348,442,500,572,648,700,1253,1346 'still':1223 'stori':1116 'strong':542 'structur':578 'studi':314,1152,1169 'suggest':945,953 'support':309,1127,1190 'surfac':73,1099 'synthes':3,48,85,502,1009,1415 'synthesi':25,136,199,511,576,635,683,727,772,832,976,1046,1059,1098,1160,1221,1328,1372,1382 'tabl':246 'take':159 'target':686,1366 'team':57,1082 'tell':121 'theme':471 'three':522 'tighter':192 'time':254,261,850 'tool':707,1016,1408 'topic':190,228,374,636,833,908,1212 'topic-agent-skills' 'topic-ai-automation' 'topic-ai-skills' 'topic-anthropic-claude' 'topic-claude' 'topic-claude-code-plugin' 'topic-claude-code-skills' 'topic-claude-cowork' 'topic-claude-desktop' 'topic-claude-skills' 'topic-cowork' 'topic-knowledge-work' 'total':903 'tri':1157 'troubleshoot':1191 'true':125 'trust':1266 'turn':680 'two':537,1295 'type':282,287,335,359,380,398,409,425,430,435,440,759,988,1007,1070,1129,1280 'under-research':1217 'use':33,197,369,989,1239,1339,1380 'user':91,145,625,925,966,1006,1069,1105,1128,1181 'vagu':346 'valuabl':1224 'verifi':744 'video':1015,1061 'vs':126,259,487,1078 'want':664 'weak':1178 'web':708 'websit':300,323,1087 'week':1140,1394 'weekly-business-puls':1393 'whether':866 'without':556,688,1368 'work':1139 'worker':60,64,164,181,337,352,371,377,383,411,422,427,432,437,452,674,722,764,766,997,1026,1084,1255,1359 'write':266,353,508 'yet':419 'yield':751","prices":[{"id":"6aad38aa-c3b9-430e-9713-6c1083bf6afc","listingId":"7c54a423-5940-486f-a838-3eeb00c9b304","amountUsd":"0","unit":"free","nativeCurrency":null,"nativeAmount":null,"chain":null,"payTo":null,"paymentMethod":"skill-free","isPrimary":true,"details":{"org":"EAIconsulting","category":"cowork-skills-library","install_from":"skills.sh"},"createdAt":"2026-05-18T13:21:05.771Z"}],"sources":[{"listingId":"7c54a423-5940-486f-a838-3eeb00c9b304","source":"github","sourceId":"EAIconsulting/cowork-skills-library/research-synthesizer","sourceUrl":"https://github.com/EAIconsulting/cowork-skills-library/tree/main/skills/research-synthesizer","isPrimary":false,"firstSeenAt":"2026-05-18T13:21:05.771Z","lastSeenAt":"2026-05-18T19:13:34.542Z"}],"details":{"listingId":"7c54a423-5940-486f-a838-3eeb00c9b304","quickStartSnippet":null,"exampleRequest":null,"exampleResponse":null,"schema":null,"openapiUrl":null,"agentsTxtUrl":null,"citations":[],"useCases":[],"bestFor":[],"notFor":[],"kindDetails":{"org":"EAIconsulting","slug":"research-synthesizer","github":{"repo":"EAIconsulting/cowork-skills-library","stars":7,"topics":["agent-skills","ai-automation","ai-skills","anthropic-claude","claude","claude-code-plugin","claude-code-skills","claude-cowork","claude-desktop","claude-skills","cowork","knowledge-work","no-code","productivity","skill-library"],"license":null,"html_url":"https://github.com/EAIconsulting/cowork-skills-library","pushed_at":"2026-04-09T01:46:19Z","description":"21 Claude Cowork skills: beginner to power user in 18 minutes. Free forever. By Everyday AI.","skill_md_sha":"0c480852fc35f58cd6bf9b69f4953648d0b5af94","skill_md_path":"skills/research-synthesizer/SKILL.md","default_branch":"main","skill_tree_url":"https://github.com/EAIconsulting/cowork-skills-library/tree/main/skills/research-synthesizer"},"layout":"multi","source":"github","category":"cowork-skills-library","frontmatter":{"name":"research-synthesizer","description":"Gathers information from multiple sources in parallel, compares what they say, flags where they agree or conflict, and delivers a clean synthesis with a confidence rating for each finding. Useful for making evidence-backed decisions, when evaluating competing claims, during due diligence on a vendor or market, and for building research briefs before strategy sessions. Use when the user says \"research this\", \"synthesize multiple sources\", \"compare sources\", \"research report\", \"what does the evidence say\", \"gather information on\", or \"deep dive into\"."},"skills_sh_url":"https://skills.sh/EAIconsulting/cowork-skills-library/research-synthesizer"},"updatedAt":"2026-05-18T19:13:34.542Z"}}