Datasets:
Tasks:
Text Classification
Formats:
csv
Languages:
English
Size:
10K - 100K
Tags:
context-engineering
loop-engineering
harness-engineering
llm-agents
systematic-review
literature-review
License:
File size: 6,938 Bytes
c8b99a6 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 | {
"exclude_section": "pre-training|post-training|prompt engineering|ontology engineering|courses|tutorials|social media|related awesome|adjacent collections|^templates$|graph engineering$",
"architecture": "attention|positional|\\brope\\b|transformer architecture|kv[- ]?cache|state[- ]space|mamba|linear attention|sparse attention|position interpolation|length extrapolation|tokeni[sz]|speculative decoding|quantiz|pre-?train",
"in_scope": "agent|context|memory|retriev|\\brag\\b|tool|harness|loop|orchestrat|multi-agent|plan|reflect|verif|critic|refine|search|workflow|skill|\\bmcp\\b|sandbox|injection|compress|compaction|summar|benchmark|bench\\b|long-horizon|prompt|reason|llm|language model|coding|software|protocol|eval",
"context_operations": {
"Compose": "system prompt|instruction file|instructions? (file|artifact)|agents\\.md|claude\\.md|few-shot|in-context learning|prompt (template|design|cach)|prompt caching|kv reuse|\\bprefix\\b|spec-driven|specification|cursor rules|project memory|context file",
"Select": "retriev|\\brag\\b|graphrag|rerank|search engine|semantic search|code search|knowledge graph|tool (selection|retrieval)|skill|\\bmcp\\b|just-in-time|progressive disclosure|\\brecall\\b|grounding|knowledge (base|retrieval)|indexing|embedding",
"Write": "memory|memories|scratchpad|\\bnotes?\\b|notebook|state file|progress (file|log)|persist|working state|journal|ledger|wiki|knowledge store|episodic|long-term",
"Compress": "compress|compaction|\\bcompact|summari|prun|truncat|condens|observation mask|masking|lingua|folding|token (budget|reduction|efficien)|context (window )?management",
"Isolate": "sub-?agents?|isolat|separate context|context folding|recursive language model|delegat|fork(ed|ing)? (context|agent)|parallel agents",
"Evolve": "self-evolv|evolving context|evolv(e|ing) (prompt|playbook|context)|dspy|gepa|playbook|cheatsheet|prompt optimi|context optimi|learn(ed|s|ing)? to (retrieve|manage|compress|curate)|curation|experiential|experience (replay|library)|agentic context engineering|evolv\\w* (agent )?skills?",
"Guard": "injection|provenance|untrusted|guardrail|poison|privacy|jailbreak|exfiltrat|secret|data leak|trust boundar|context integrity"
},
"loop_levels": {
"L1 Action": "\\breact\\b|tool use|tool[- ]call|function[- ]call|agent loop|agentic loop|codeact|computer use|browser use|executable code action|act(ing)? in",
"L2 Plan": "\\bplan(s|ning|ner)?\\b|decompos|task graph|workflow|stateflow|replan|to-?do",
"L3 Verification": "verif|critic|critique|reflect|self-refine|\\brefine|evaluator|llm-as-a-judge|judge|quality gate|self-correct|proof|evidence-gated|\\btests?\\b|testing|code review|review loop|validator",
"L4 Search": "tree[- ]search|mcts|monte carlo|tree of thoughts|graph of thoughts|beam search|best-of-n|repeated sampling|\\blats\\b|search over|test-time (compute|scaling)|branch",
"L5 Orchestration": "multi-agent|orchestrat|coordinat|controller|supervisor|swarm|\\bteams?\\b|handoff|router|manager agent|worker",
"L6 Session": "long-running|long-horizon|multi-session|across sessions|context windows|\\bsession|resum|initializer|ralph|persistent agent|durable execution|checkpoint",
"L7 Scheduled": "schedul|\\bcron\\b|background agent|event-driven|trigger|daemon|unattended|outer loop|loop engineering|triage|always-on|ambient|autonomous (loop|agents? that run)",
"L8 Improvement": "self-improv|self-evolv|evolv|meta-harness|harness optimi|automated design|\\badas\\b|aflow|darwin|alphaevolve|funsearch|self-modif|(prompt|workflow|agent|harness) optimi[sz]ation|self-harness|automatic evolution"
},
"harness": "harness|runtime|sandbox|permission|authori[sz]|tool design|tool interface|protocol|\\bmcp\\b|execution|scaffold|observability|tracing|telemetry|infrastructure|container|\\bsdk\\b|framework|hooks?\\b|cli\\b|desktop|control plane",
"evaluation": "benchmark|\\bbench\\b|-bench|arena|leaderboard|\\bdataset|evaluat|\\beval",
"section_layer": [
[
"harness",
"runtime orchestration|interaction paradigm|agent loop|harness architecture|tool integration|reference harness|coding-agent harness|execution substrates|runtime foundations"
],
[
"evaluation",
"benchmark|\\bevals?\\b|\\bevaluation|benchmarking|quality assessment"
],
[
"context",
"context|memory|retrieval|\\brag\\b|knowledge|structured data|self-generated|compaction|working-state|instruction files|project memory|skill composition"
],
[
"loop",
"agent loop|loop architecture|planning|orchestration|verification|human-in-the-loop|task runners|multi-agent|agent coordination|task organization|system evolution|environment feedback|interaction paradigm|state management|spec-driven|operating principles|operational autonomy|meta-harness|generators|workflow, profiles"
],
[
"harness",
"harness|runtime|sandbox|security|permission|tool design|tool integration|skills|protocols|observability|debugging|production|execution|guardrails|reference|generators|tool use|function calling|agent communication|skill composition|foundations|platform|implementations|telemetry|browser, mcp|workflow, profiles|constraints"
]
],
"section_context_ops": {
"Select": "retrieval|\\brag\\b|knowledge|structured data|tool integration|skills",
"Write": "memory|state management|working-state|project memory",
"Compress": "compression|compaction|context management in production",
"Compose": "instruction files|context design|context delivery|prompt",
"Guard": "security|guardrails|permissions"
},
"section_loop_levels": {
"L1": "agent loop|tool use|function calling",
"L2": "planning|task organization|spec-driven|task decomposition",
"L3": "verification|quality gates|environment feedback",
"L5": "orchestration|multi-agent|agent coordination|task runners",
"L8": "system evolution|generators & meta-harnesses"
},
"contribution": [
[
"survey",
"\\bsurvey|systematic review|\\breview\\b|taxonomy|overview|position paper|perspective|\\bsok\\b|primer"
],
[
"benchmark",
"benchmark|\\bbench\\b|-bench|arena|leaderboard|\\bdataset|evaluat|\\beval"
],
[
"empirical",
"empirical|exploratory|\\bstudy\\b|analysis of|we find|measur|understanding|how (do|does|well)|lessons|investigat|uncovering|diagnos"
]
],
"domain": [
[
"software",
"\\bcode\\b|coding|software|\\bswe\\b|repositor|programming|developer|pull request|terminal|\\bci\\b|github|compiler|debug"
],
[
"web_gui",
"\\bweb\\b|browser|\\bgui\\b|computer use|desktop|mobile|android|\\bos\\b|screen"
],
[
"research",
"research|scien|discovery|literature|paper|chemistry|biolog|medical|clinical"
],
[
"dialogue",
"conversation|dialog|personal|assistant|chatbot|user model|companion|customer"
]
]
} |