Pending content
1979:No items declared.
1980: ${phase.summary}
1404: `; 1405: btn.title = phase.summary; 1406: btn.addEventListener("click", () => { 1407: state.phaseId = phase.id; 1408: const matchingCategory = HUB.categories.find(cat => cat.docs.some(doc => doc.phase === phase.id)); 1409: if (matchingCategory) { 1410: state.categoryId = matchingCategory.id; 1411: state.docId = matchingCategory.docs[0].id; ===== SLICE AROUND renderArchitectureView ===== 1956: previewItems.length 1957: ? `${escapeHtml("Work Policy Center Canonical is the rule cabinet for DEVON-DEV execution. It collects the operating policies that cannot live as scattered prompt memory: project instructions, coding standards, patch rules, sandbox rules, production gates, documentation discipline, planning requirements, validation policy, rollback policy, security boundaries, memory behavior, learning limits, UI conduct and server operation rules.")}
1991:${escapeHtml("FUNCTION: Give the future Execution & Monitoring Panel a single policy authority to consult before work becomes executable. The panel must not allow a plan, tool call, patch, shell command, sandbox action, production promotion, memory write, learning update or UI operation when the applicable work policy is absent, unresolved or contradictory.")}
1995:${escapeHtml("This category does not decide individual actions; it supplies the policy corpus that Decision Rule Registry and Policy Compiler turn into enforceable FSM rules. Documentation Before Code consumes its documentation policy, Requirements & Scope Engine consumes its planning boundary, Tool Registry consumes its operation rules, Validation Matrix consumes its validation policy, and Artifact Registry consumes its rollback and evidence requirements.")}
1996:${escapeHtml("A DEVON-DEV workflow without this center drifts back into chat-driven behavior: each task depends on whatever the prompt remembered at that moment. Missing policy is MISSING. Conflicting policy is FAIL. A policy that cannot be compiled into a gate, block, evidence requirement, approval rule or recovery path is not operational governance; it is documentation noise.")}
1997:${escapeHtml("Local LLM Runtime Canonical fixes the execution boundary for DEVON-DEV interpretation. The system may use model reasoning, code analysis and operational reading, but that interpretive layer cannot be based on an invisible external API, an assumed cloud dependency or a model runtime that has not been installed, benchmarked and exposed to the future panel.")}
36883:${escapeHtml("FUNCTION: Establish the local or self-hosted runtime as the only acceptable base for LLM execution while keeping the final runtime product MISSING until benchmark evidence exists. This gives the future Execution & Monitoring Panel a concrete runtime target to inspect, not a vague statement that a model will be available somewhere.")}
36887:${escapeHtml("This authority is consumed before Qwen Coder Model, Qwen Data Interpreter, Internal RAG, Prompt Governance and ModelOps can claim operational readiness. It separates interpretation from decision: the runtime can host the model, the model can interpret, but FSM Absolute Decision and Decision Rule Registry remain responsible for permission, gates and promotion.")}
36888:${escapeHtml("MISSING applies while no local runtime, container target, benchmark result or validated execution path exists. FAIL applies if DEVON-DEV silently depends on an external API, treats GPU as mandatory without an explicit decision, promotes a runtime without benchmark, or lets model availability masquerade as system authority. Without this category, the entire cognitive layer rests on assumption instead of inspectable infrastructure.")}
36889:${escapeHtml("CFC is the sovereign flow law for Phase 03. It decides how cognitive work is allowed to move from context intake to role interaction, evidence preparation, runtime readiness and downstream consumption without letting reasoning order become a hidden improvisation inside the assistant or the future panel.")}
36753:${escapeHtml("FUNCTION: Enforce deterministic cognitive-flow sequencing before runtime, retrieval, model interpretation, prompt assembly or monitoring layers consume Phase 03. This category gives the future Execution & Monitoring Panel a single flow authority to inspect when it needs to know what may receive context, what may transform it, what must wait for evidence and what must remain blocked from probabilistic shortcut behavior.")}
36757:${escapeHtml("CFC is consumed by Local LLM Runtime, Qwen model authorities, Internal RAG, Retrieval Engine, Embeddings Engine, Vector Index, Prompt Governance, Prompt / Instruction Assembly and the later planning and focus controls. It does not choose a model, execute a tool or approve a plan; it preserves the lawful order that those categories must obey before they claim cognitive readiness.")}
36758:${escapeHtml("MISSING applies when Phase 03 has components but no declared flow law. FAIL applies when a downstream category invents its own order, receives context before evidence is prepared, bypasses flow state, or lets runtime availability replace deterministic orchestration. Without this category, DEVON-DEV can assemble cognition from valid parts and still operate as an unordered, non-auditable reasoning chain.")}
36759: