[{"data":1,"prerenderedAt":114},["ShallowReactive",2],{"\u002Fdocs\u002Fcore-concepts\u002Flayer-1":3},{"id":4,"title":5,"body":6,"description":106,"extension":107,"meta":108,"navigation":109,"path":110,"seo":111,"stem":112,"__hash__":113},"docs\u002Fdocs\u002Fcore-concepts\u002Flayer-1.md","Layer 1: Project Knowledge Base",{"type":7,"value":8,"toc":98},"minimark",[9,13,18,21,25,58,62,77,81],[10,11,12],"p",{},"Layer 1 is the foundation of CAF's five-layer architecture. It holds context about\nyour specific project — not general knowledge about a framework or language, but the\ndecisions and conventions that apply to your own repo.",[14,15,17],"h2",{"id":16},"why-this-layer-exists","Why this layer exists",[10,19,20],{},"Without a Project Knowledge Base, every agent would have to re-guess your stack,\nfolder structure, and code conventions each time it runs — leading to inconsistent\nresults across tickets, and agents that might make decisions conflicting with\nexisting patterns in the repo. Layer 1 ensures every agent — Planner, Implementer,\nVerifier, Reviewer — reads the same context before starting work.",[14,22,24],{"id":23},"what-it-contains","What it contains",[26,27,28,40,46,52],"ul",{},[29,30,31,35,36],"li",{},[32,33,34],"strong",{},"Stack & tooling"," — framework, package manager, runtime version, automatically\ndetected by CAF Initiator during ",[37,38,39],"code",{},"init",[29,41,42,45],{},[32,43,44],{},"Folder structure"," — the file placement conventions already in place in the repo",[29,47,48,51],{},[32,49,50],{},"Architecture decisions"," — deliberate constraints or patterns (e.g. \"all state\nmanagement uses X, not Y\")",[29,53,54,57],{},[32,55,56],{},"Existing code standards"," — linter\u002Fformatter rules, naming conventions already in\nuse",[14,59,61],{"id":60},"where-this-file-lives","Where this file lives",[10,63,64,65,68,69,72,73,76],{},"CAF Initiator generates this knowledge base as ",[37,66,67],{},".claude\u002FPROJECT.md"," when you run\n",[37,70,71],{},"caf-initiator init",". This file can be edited manually at any time — CAF won't\noverwrite it on subsequent runs unless you explicitly request it via\n",[37,74,75],{},"caf-initiator export",".",[14,78,80],{"id":79},"relationship-to-other-layers","Relationship to other layers",[10,82,83,84,89,90,93,94,97],{},"Layer 1 is the input for ",[85,86,88],"a",{"href":87},"\u002Fdocs\u002Fcore-concepts\u002Flayer-2","Layer 2: Agent Definitions"," —\nevery agent definition in ",[37,91,92],{},".claude\u002Fagents\u002F"," refers back to ",[37,95,96],{},"PROJECT.md"," so each role's\nbehavior stays aligned with the same project context.",{"title":99,"searchDepth":100,"depth":100,"links":101},"",2,[102,103,104,105],{"id":16,"depth":100,"text":17},{"id":23,"depth":100,"text":24},{"id":60,"depth":100,"text":61},{"id":79,"depth":100,"text":80},"The foundation of CAF's architecture — the project context every agent reads before acting.","md",{},true,"\u002Fdocs\u002Fcore-concepts\u002Flayer-1",{"title":5,"description":106},"docs\u002Fcore-concepts\u002Flayer-1","SmC_VSZojLNt-9pvdFhpk8aZgCnvZoV0pQYi-2oEJs8",1787621083728]