{"id":"777dc38c-c948-4771-acc5-3b7abdfd4bc7","shortId":"dk5QdM","kind":"skill","title":"synthesize","tagline":"Generate a research paper from investigation results, structured as a formal academic paper with definitions, theorems, lemmas, and proofs. Use when asked to synthesize, summarize, or write up research findings after investigation cycles.","description":"# Synthesize\n\nGenerate a formal research paper from all investigation results. The output should read like a publishable academic paper — not a consulting report or investigation diary.\n\n## Usage\n\nInvoke this skill by name; no arguments required. On slash-command hosts, prefix with `/` (e.g. `/synthesize`).\n\n```\nsynthesize\n```\n\n## Instructions\n\n### 1. Re-read Workspace State (Efficiently)\n\nDo NOT rely on context window. Re-read workspace files, but **load selectively** to stay within context limits:\n\n**Always read:**\n- `reaper-workspace/notes/current-understanding.md` — the accumulated findings (primary source)\n- `reaper-workspace/notes/results.md` — the cycle-by-cycle record including batch summaries\n- `reaper-workspace/notes/problem-statement.md` — model assumptions and property definitions\n- `reaper-workspace/notes/ideas.md` — the ideas and their resolution status\n\n**Read for background (skim, don't deep-read):**\n- `reaper-workspace/notes/paper-summary.md` — focus on \"Problem Statement\" and \"Red Flags\" sections (skip if not present)\n- `reaper-workspace/notes/literature.md` — focus on \"Landscape Summary\" and \"Key Prior Results\" sections\n\n**Load investigation directories selectively:**\n- Read `notes/results.md` batch summaries first — these provide a condensed view of each batch's findings\n- Only read `investigations/*/analysis.md` for **keep hypotheses** (check the \"Status\" column in `notes/results.md`). Skip discard investigations entirely. Each `analysis.md` reflects the latest state of its hypothesis (edited inline on revisit).\n- If batch summaries are present and sufficient, you may skip individual `analysis.md` files for cycles already summarized\n- For investigations with formal proofs (`proof.md`), only read if the report needs to cite a specific proof step\n\n### 2. Identify the Central Result\n\nIf the investigation yielded multiple results, identify the **single most important one**. This is the main theorem of the paper. Ask: \"If the reader remembers only one thing, what should it be?\"\n\n### 3. Write the Paper\n\nWrite to `reaper-workspace/report.md` following the structure of a formal academic paper. The exact sections depend on the nature of the results (new construction, impossibility, proof gap, analysis), but the general template is:\n\n```markdown\n# [Title: Descriptive, Not Generic]\n\n## Abstract\n\n[Problem, approach, main result(s), significance. State the main theorem informally here. No more than 350 words — be concise. Split into multiple short paragraphs if needed for readability.]\n\n## 1. Introduction\n\n### 1.1 Problem and Motivation\n[What problem does this address? Why does it matter?]\n\n### 1.2 Our Results\n[Informal statement of main results. Each contribution must be specific and refutable — a reader could disagree.]\n\n- **Theorem 1 (informal).** [Main result in plain language]\n- **Theorem 2 (informal).** [Secondary result, if any]\n\nNOT: \"We analyze protocol X\" or \"We study the security of Y\"\n\n### 1.3 Related Work\n[How this relates to prior work. What was known before. What gap this fills.]\n\n### 1.4 Paper Organization\n[Brief roadmap of remaining sections.]\n\n## 2. Preliminaries\n\n### 2.1 Notation\n[Notation conventions used throughout.]\n\n### 2.2 Model\n[Formal system/threat model. Every dimension pinned down — consult the problem-statement.md model assumptions. State these as a definition if appropriate.]\n\n### 2.3 Definitions\n[Formal definitions of properties, security notions, or concepts used in theorems. Use Definition environments.]\n\n**Definition 1** (Property Name). [Formal definition]\n\n## 3. [Main Technical Content]\n\nThis is the core of the paper. Structure depends on the type of result:\n\n**For new constructions or protocols:**\n- Section 3: Construction (protocol description)\n- Section 4: Security Analysis (theorems + proofs)\n- Section 5: Performance Analysis (complexity claims + proofs)\n\n**For proof gaps or attacks:**\n- Section 3: The Claimed Result (what the original paper claimed)\n- Section 4: The Gap/Attack (where and why it fails, with concrete example first)\n- Section 5: Implications and Fixes (what this means, minimal repairs if any)\n\n**For impossibility results:**\n- Section 3: The Impossibility (theorem statement)\n- Section 4: Proof of Impossibility\n- Section 5: Circumventing the Impossibility (what additional assumptions restore possibility)\n\n**For comparative analysis:**\n- Section 3: Framework (dimensions of comparison)\n- Section 4: Analysis of Each Approach (with formal claims)\n- Section 5: Comparison (theorems about relative strengths/weaknesses)\n\nWithin each section, organize results as a logical chain of definitions, lemmas, theorems, and proofs:\n\n**Lemma 1.** [Precise statement with all assumptions explicit]\n\n*Proof.* [Step-by-step argument. Each step justified by definition, assumption, or prior lemma. End with ∎]\n\n**Theorem 1.** [Main result — precise mathematical statement]\n\n*Proof.* [May reference lemmas. Proof technique stated. Complete argument.]\n\nFor counterexamples and attacks, use a concrete execution trace:\n\n**Claim 1.** [The property that fails]\n\n*Counterexample.* Consider n = [specific value], with adversary strategy... [Concrete execution trace showing the violation.] ∎\n\n## 4. Discussion\n\n[Implications, limitations, comparison to prior understanding. What assumptions could be relaxed or strengthened.]\n\n## 5. Open Questions\n\n[Specific, actionable open problems — not \"more work is needed.\" State each as a concrete question or conjecture.]\n\n**Conjecture 1.** [Precise statement of what you believe but did not prove]\n\n## References\n\n[Papers cited during the investigation, formatted consistently. For each paper, include the **publication venue** (e.g., CRYPTO 2023, S&P 2024, J. Cryptology Vol 36) — not just an arXiv or ePrint ID. The venue should already be resolved in `notes/literature.md`'s `Venue` column from the `/review-literature` step; copy it through. For papers without a known venue, label as `(preprint)` and include the archive URL. Also attach a link to the PDF (arXiv, IACR ePrint, DOI, or publisher URL).\n\nExample formats:\n- Author, A., Author, B. *Title.* CRYPTO 2023, pp. XXX–YYY. [arXiv:XXXX.XXXXX](https://arxiv.org/abs/XXXX.XXXXX)\n- Author, C. *Title.* (preprint) IACR ePrint 2024/1234. [ePrint](https://eprint.iacr.org/2024/1234)]\n\n## Appendix A: Investigation Log\n\n[Summary table from notes/results.md — the full cycle-by-cycle record for reproducibility.]\n```\n\n### 4. Writing Principles\n\n**Structure of a formal paper, not a report:**\n- Every claim is backed by a definition, theorem, lemma, or proof — not just prose assertions\n- Concrete examples (execution traces, adversary strategies) appear *before* general theorems, motivating the formal result\n- The logical chain must be self-contained: a reader should be able to verify each proof step from the definitions and assumptions stated in the paper\n\n**Peyton Jones principles still apply:**\n- **One \"ping\"**: the paper has one clear main theorem stated upfront in the abstract and introduction\n- **Every claim is refutable**: if a sentence could appear in any paper about any topic, it's too vague — cut it or formalize it\n- **Problem → why it matters → what we found → evidence → comparison** — NOT a chronological diary of the investigation\n\n**Formal rigor:**\n- Every theorem/lemma states its assumptions explicitly\n- Proofs cite the proof technique used (reduction, simulation, induction, game-hopping, etc.)\n- Distinguish between what is proven and what is conjectured — label conjectures clearly\n- If a proof has gaps (e.g., a step that relies on intuition), state this honestly as a remark\n\n### Confidence Handling\n\nPreserve confidence levels from `notes/results.md` exactly as logged — do not upgrade. The investigate skill defaults one level lower than instinct; synthesize must not undo that calibration. Map to paper constructs: **high** → theorems with full proofs, **medium** → propositions with explicit gap statements, **low** → clearly labeled conjectures.\n\n### Failure Recovery\n\nIf `current-understanding.md` is empty or missing, reconstruct from `results.md` batch summaries and `investigations/*/analysis.md` before writing. If both are empty, return an error — there is nothing to synthesize.\n\n### Quality Criteria\n\n- Main result is stated as a formal theorem in the introduction (informal) and body (formal)\n- Every non-trivial claim has a corresponding theorem/lemma with proof\n- Definitions are precise — no informal \"safety\" or \"liveness\" without formal predicate\n- Counterexamples include concrete execution traces, not just \"this can fail\"\n- The paper is self-contained — a reader unfamiliar with the investigation can follow the argument\n- No chronological narration (\"first we tried X, then we tried Y\") — present the clean logical chain\n- Open questions are stated as conjectures or precise problems, not vague future work\n- Every entry in the References section names a real publication venue or is explicitly labeled `(preprint)` — never just an arXiv/ePrint ID where a venue is expected","tags":["synthesize","reaper","sebastianelvis","agent-skills","ai-research","arxiv","claude-code","cline","codex-cli","cryptography","cursor","dblp"],"capabilities":["skill","source-sebastianelvis","skill-synthesize","topic-agent-skills","topic-ai-research","topic-arxiv","topic-claude-code","topic-cline","topic-codex-cli","topic-cryptography","topic-cursor","topic-dblp","topic-distributed-systems","topic-gemini-cli","topic-iacr"],"categories":["reaper"],"synonyms":[],"warnings":[],"endpointUrl":"https://skills.sh/SebastianElvis/reaper/synthesize","protocol":"skill","transport":"skills-sh","auth":{"type":"none","details":{"cli":"npx skills add SebastianElvis/reaper","source_repo":"https://github.com/SebastianElvis/reaper","install_from":"skills.sh"}},"qualityScore":"0.454","qualityRationale":"deterministic score 0.45 from registry signals: · indexed on github topic:agent-skills · 8 github stars · SKILL.md body (9,041 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":"2026-05-18T19:08:57.539Z","embedding":null,"createdAt":"2026-05-18T13:14:42.078Z","updatedAt":"2026-05-18T19:08:57.539Z","lastSeenAt":"2026-05-18T19:08:57.539Z","tsv":"'/2024/1234)]':886 '/abs/xxxx.xxxxx)':875 '/analysis.md':210,1145 '/notes/current-understanding.md':112 '/notes/ideas.md':143 '/notes/literature.md':178 '/notes/paper-summary.md':162 '/notes/problem-statement.md':134 '/notes/results.md':121 '/report.md':318 '/review-literature':826 '/synthesize':78 '1':81,382,417,514,665,690,715,770 '1.1':384 '1.2':397 '1.3':443 '1.4':460 '2':272,425,468 '2.1':470 '2.2':476 '2.3':497 '2023':798,867 '2024':801 '2024/1234':882 '3':309,519,543,566,604,628 '350':369 '36':805 '4':548,576,610,634,734,904 '5':554,589,615,643,749 'abl':956 'abstract':353,989 'academ':13,52,325 'accumul':114 'action':753 'addit':620 'address':392 'adversari':726,934 'alreadi':252,816 'also':845 'alway':107 'analysi':342,550,556,626,635 'analysis.md':225,248 'analyz':433 'appear':936,1000 'appendix':887 'appli':975 'approach':355,638 'appropri':496 'archiv':843 'argument':68,677,704,1224 'arxiv':809,852,871 'arxiv.org':874 'arxiv.org/abs/xxxx.xxxxx)':873 'arxiv/eprint':1273 'ask':23,297 'assert':929 'assumpt':136,489,621,670,683,743,966,1038 'attach':846 'attack':564,708 'author':861,863,876 'b':864 'back':918 'background':152 'batch':129,194,204,238,1141 'believ':776 'bodi':1175 'brief':463 'c':877 'calibr':1110 'central':275 'chain':657,946,1240 'check':214 'chronolog':1027,1226 'circumv':616 'cite':267,783,1041 'claim':558,568,574,641,714,916,993,1181 'clean':1238 'clear':982,1064,1127 'column':217,823 'command':73 'compar':625 'comparison':632,644,738,1024 'complet':703 'complex':557 'concept':506 'concis':372 'concret':585,711,728,765,930,1201 'condens':200 'confid':1083,1086 'conjectur':768,769,1061,1063,1129,1246 'consid':721 'consist':788 'construct':338,539,544,1114 'consult':56,485 'contain':951,1214 'content':522 'context':92,105 'contribut':406 'convent':473 'copi':828 'core':526 'correspond':1184 'could':414,744,999 'counterexampl':706,720,1199 'criteria':1161 'crypto':797,866 'cryptolog':803 'current-understanding.md':1133 'cut':1011 'cycl':34,124,126,251,898,900 'cycle-by-cycl':123,897 'deep':157 'deep-read':156 'default':1099 'definit':16,139,494,498,500,511,513,518,659,682,921,964,1188 'depend':330,531 'descript':350,546 'diari':60,1028 'dimens':482,630 'directori':190 'disagre':415 'discard':221 'discuss':735 'distinguish':1053 'doi':855 'e.g':77,796,1070 'edit':233 'effici':87 'empti':1135,1151 'end':687 'entir':223 'entri':1255 'environ':512 'eprint':811,854,881,883 'eprint.iacr.org':885 'eprint.iacr.org/2024/1234)]':884 'error':1154 'etc':1052 'everi':481,915,992,1034,1177,1254 'evid':1023 'exact':328,1090 'exampl':586,859,931 'execut':712,729,932,1202 'expect':1279 'explicit':671,1039,1123,1267 'fail':583,719,1208 'failur':1130 'file':98,249 'fill':459 'find':31,115,206 'first':196,587,1228 'fix':592 'flag':169 'focus':163,179 'follow':319,1222 'formal':12,38,257,324,478,499,517,640,910,942,1014,1032,1168,1176,1197 'format':787,860 'found':1022 'framework':629 'full':896,1118 'futur':1252 'game':1050 'game-hop':1049 'gap':341,457,562,1069,1124 'gap/attack':578 'general':345,938 'generat':2,36 'generic':352 'handl':1084 'high':1115 'honest':1079 'hop':1051 'host':74 'hypothes':213 'hypothesi':232 'iacr':853,880 'id':812,1274 'idea':145 'identifi':273,283 'implic':590,736 'import':287 'imposs':339,601,606,613,618 'includ':128,792,841,1200 'individu':247 'induct':1048 'inform':364,400,418,426,1173,1192 'inlin':234 'instinct':1104 'instruct':80 'introduct':383,991,1172 'intuit':1076 'investig':7,33,43,59,189,209,222,255,279,786,889,1031,1097,1144,1220 'invok':62 'j':802 'jone':972 'justifi':680 'keep':212 'key':184 'known':454,835 'label':837,1062,1128,1268 'landscap':181 'languag':423 'latest':228 'lemma':18,660,664,686,699,923 'level':1087,1101 'like':49 'limit':106,737 'link':848 'live':1195 'load':100,188 'log':890,1092 'logic':656,945,1239 'low':1126 'lower':1102 'main':292,356,362,403,419,520,691,983,1162 'map':1111 'markdown':348 'mathemat':694 'matter':396,1019 'may':245,697 'mean':595 'medium':1120 'minim':596 'miss':1137 'model':135,477,480,488 'motiv':387,940 'multipl':281,375 'must':407,947,1106 'n':722 'name':66,516,1260 'narrat':1227 'natur':333 'need':265,379,760 'never':1270 'new':337,538 'non':1179 'non-trivi':1178 'notat':471,472 'notes/literature.md':820 'notes/results.md':193,219,894,1089 'noth':1157 'notion':504 'one':288,303,976,981,1100 'open':750,754,1241 'organ':462,652 'origin':572 'output':46 'p':800 'paper':5,14,40,53,296,312,326,461,529,573,782,791,832,911,970,979,1003,1113,1210 'paragraph':377 'pdf':851 'perform':555 'peyton':971 'pin':483 'ping':977 'plain':422 'possibl':623 'pp':868 'precis':666,693,771,1190,1248 'predic':1198 'prefix':75 'preliminari':469 'preprint':839,879,1269 'present':174,241,1236 'preserv':1085 'primari':116 'principl':906,973 'prior':185,450,685,740 'problem':165,354,385,389,755,1016,1249 'problem-statement.md':487 'proof':20,258,270,340,552,559,561,611,663,672,696,700,925,960,1040,1043,1067,1119,1187 'proof.md':259 'properti':138,502,515,717 'proposit':1121 'prose':928 'protocol':434,541,545 'prove':780 'proven':1057 'provid':198 'public':794,1263 'publish':51,857 'qualiti':1160 'question':751,766,1242 're':83,95 're-read':82,94 'read':48,84,96,108,150,158,192,208,261 'readabl':381 'reader':300,413,953,1216 'real':1262 'reaper':110,119,132,141,160,176,316 'reaper-workspac':109,118,131,140,159,175,315 'reconstruct':1138 'record':127,901 'recoveri':1131 'red':168 'reduct':1046 'refer':698,781,1258 'reflect':226 'refut':411,995 'relat':444,448,647 'relax':746 'reli':90,1074 'remain':466 'remark':1082 'rememb':301 'repair':597 'report':57,264,914 'reproduc':903 'requir':69 'research':4,30,39 'resolut':148 'resolv':818 'restor':622 'result':8,44,186,276,282,336,357,399,404,420,428,536,569,602,653,692,943,1163 'results.md':1140 'return':1152 'revisit':236 'rigor':1033 'roadmap':464 'safeti':1193 'secondari':427 'section':170,187,329,467,542,547,553,565,575,588,603,609,614,627,633,642,651,1259 'secur':440,503,549 'select':101,191 'self':950,1213 'self-contain':949,1212 'sentenc':998 'short':376 'show':731 'signific':359 'simul':1047 'singl':285 'skill':64,1098 'skill-synthesize' 'skim':153 'skip':171,220,246 'slash':72 'slash-command':71 'sourc':117 'source-sebastianelvis' 'specif':269,409,723,752 'split':373 'state':86,229,360,490,702,761,967,985,1036,1077,1165,1244 'statement':166,401,608,667,695,772,1125 'status':149,216 'stay':103 'step':271,674,676,679,827,961,1072 'step-by-step':673 'still':974 'strategi':727,935 'strengthen':748 'strengths/weaknesses':648 'structur':9,321,530,907 'studi':438 'suffici':243 'summar':26,253 'summari':130,182,195,239,891,1142 'synthes':1,25,35,79,1105,1159 'system/threat':479 'tabl':892 'technic':521 'techniqu':701,1044 'templat':346 'theorem':17,293,363,416,424,509,551,607,645,661,689,922,939,984,1116,1169 'theorem/lemma':1035,1185 'thing':304 'throughout':475 'titl':349,865,878 'topic':1006 'topic-agent-skills' 'topic-ai-research' 'topic-arxiv' 'topic-claude-code' 'topic-cline' 'topic-codex-cli' 'topic-cryptography' 'topic-cursor' 'topic-dblp' 'topic-distributed-systems' 'topic-gemini-cli' 'topic-iacr' 'trace':713,730,933,1203 'tri':1230,1234 'trivial':1180 'type':534 'understand':741 'undo':1108 'unfamiliar':1217 'upfront':986 'upgrad':1095 'url':844,858 'usag':61 'use':21,474,507,510,709,1045 'vagu':1010,1251 'valu':724 'venu':795,814,822,836,1264,1277 'verifi':958 'view':201 'violat':733 'vol':804 'window':93 'within':104,649 'without':833,1196 'word':370 'work':445,451,758,1253 'workspac':85,97,111,120,133,142,161,177,317 'write':28,310,313,905,1147 'x':435,1231 'xxx':869 'xxxx.xxxxx':872 'y':442,1235 'yield':280 'yyy':870","prices":[{"id":"05a9ad91-bcb6-475f-a365-e4f99ecc5156","listingId":"777dc38c-c948-4771-acc5-3b7abdfd4bc7","amountUsd":"0","unit":"free","nativeCurrency":null,"nativeAmount":null,"chain":null,"payTo":null,"paymentMethod":"skill-free","isPrimary":true,"details":{"org":"SebastianElvis","category":"reaper","install_from":"skills.sh"},"createdAt":"2026-05-18T13:14:42.078Z"}],"sources":[{"listingId":"777dc38c-c948-4771-acc5-3b7abdfd4bc7","source":"github","sourceId":"SebastianElvis/reaper/synthesize","sourceUrl":"https://github.com/SebastianElvis/reaper/tree/main/skills/synthesize","isPrimary":false,"firstSeenAt":"2026-05-18T13:14:42.078Z","lastSeenAt":"2026-05-18T19:08:57.539Z"}],"details":{"listingId":"777dc38c-c948-4771-acc5-3b7abdfd4bc7","quickStartSnippet":null,"exampleRequest":null,"exampleResponse":null,"schema":null,"openapiUrl":null,"agentsTxtUrl":null,"citations":[],"useCases":[],"bestFor":[],"notFor":[],"kindDetails":{"org":"SebastianElvis","slug":"synthesize","github":{"repo":"SebastianElvis/reaper","stars":8,"topics":["agent-skills","ai-research","arxiv","claude-code","cline","codex-cli","cryptography","cursor","dblp","distributed-systems","gemini-cli","iacr","openalex","research","scientific-research","semantic-scholar","skills"],"license":"apache-2.0","html_url":"https://github.com/SebastianElvis/reaper","pushed_at":"2026-05-02T06:06:07Z","description":"AI-native scientific research pipeline. Takes a research goal and autonomously runs multi-step academic research on Cursor, Codex CLI, Cline, Continue, Gemini CLI, Copilot, Windsurf, Claude Code, and 40+ more.","skill_md_sha":"4d560ff11f7cb55e0d7461ef6902bd6f183445d2","skill_md_path":"skills/synthesize/SKILL.md","default_branch":"main","skill_tree_url":"https://github.com/SebastianElvis/reaper/tree/main/skills/synthesize"},"layout":"multi","source":"github","category":"reaper","frontmatter":{"name":"synthesize","license":"Apache-2.0","description":"Generate a research paper from investigation results, structured as a formal academic paper with definitions, theorems, lemmas, and proofs. Use when asked to synthesize, summarize, or write up research findings after investigation cycles."},"skills_sh_url":"https://skills.sh/SebastianElvis/reaper/synthesize"},"updatedAt":"2026-05-18T19:08:57.539Z"}}