# The algo records whose force is model_governance

This index as JSON: https://banes-lab.com/json/api/facets/algo/force/model-governance

## Entries

- [Evidence-Before-Generation](https://banes-lab.com/records/algo/evidence-before-generation.md): Discover available context, inspect target-domain resources, extract concrete facts, construct a knowledge base, and generate only from verified evidence.
- [Phase Validation Requirement](https://banes-lab.com/records/algo/phase-validation-requirement.md): For every phase boundary, generate explicit validation requirements from the phase purpose, adapter-separation needs, evidence-grounding needs, and safety constraints.
- [Validation Strategy Composition](https://banes-lab.com/records/algo/validation-strategy-composition.md): Define pre-generation, during-generation, and post-generation checks that enforce capabilities, creation history, evidence grounding, adapter separation, schema validity, and absence of unsupported assumptions.
- [Evidence Grounding Validation](https://banes-lab.com/records/algo/evidence-grounding-validation.md): Search generated output for unsupported claims and evidence references, compare generated claims against the knowledge base, calculate grounding score, and reject artifacts below threshold.
- [Agent Generation Completion](https://banes-lab.com/records/algo/agent-generation-completion.md): Accept the generated agent as complete only when semantic-compliance, evidence-grounding above threshold, and algorithmic-embodiment all pass and the audit trail is recorded; otherwise route to regeneration or a blocked report.
- [<Agent Generation Concern>](https://banes-lab.com/records/algo/agent-generation-concern.md): <Load context> → <Detect capabilities> → <Check collisions> → <Discover domain evidence> → <Build knowledge> → <Analyze risk> → <Design phases> → <Compose portable contract> → <Render adapters> → <Validate grounding> → <Audit> → <Report>
- [Hybrid Workflow Orchestration](https://banes-lab.com/records/algo/hybrid-workflow-orchestration.md): Classify workflow phases by execution verb, run discovery and investigation agents in parallel forked contexts, run action and mutation agents sequentially, and coordinate all agents through shared artifacts and handoff signals.
- [DSL Compliance Loading](https://banes-lab.com/records/algo/dsl-compliance-loading.md): Load grammar, keyword, operator, workflow, and checklist references before workflow generation, then require generated content to follow the declared DSL, naming, handoff, execution, forking, spawning, and capability rules.
- [Context Forking Configuration](https://banes-lab.com/records/algo/context-forking-configuration.md): Assign forked context to discovery, investigation, and documentation agents; assign normal context to action and validation agents; preserve model and parallel eligibility metadata per agent class.
- [Verb-Based Execution Classification](https://banes-lab.com/records/algo/verb-based-execution-classification.md): Extract the primary verb from each agent purpose, classify it against parallel or sequential verb sets, and bind execution mode, context, and parallel eligibility.
- [Workspace Configuration Discovery](https://banes-lab.com/records/algo/workspace-configuration-discovery.md): Read workspace configuration, extract zones, semantic extensions, root path, agent definitions, shared zone, workflow zone, and task zone before generating any workflow artifacts.
- [Shared Document Workspace](https://banes-lab.com/records/algo/shared-document-workspace.md): Pre-create workflow documents once, require every agent to read and edit the same documents, prohibit duplicate versions, and enforce single source of truth across phases.
- [Workflow Type Document Selection](https://banes-lab.com/records/algo/workflow-type-document-selection.md): Analyze workflow objective, classify workflow type, select five to eight semantically distinct core documents, and map each document to a non-overlapping purpose.
- [Agent Sequence Definition](https://banes-lab.com/records/algo/agent-sequence-definition.md): For each agent slot, create an agent object with name, type, phase, purpose, methodology, inputs, outputs, validation gates, document responsibilities, focus areas, edit protocol, agent activation, and 4D graph.
- [Agent Document Responsibility](https://banes-lab.com/records/algo/agent-document-responsibility.md): For each core document, require every agent to read current state, identify outdated or incorrect content, remove stale sections, replace with new discoveries, avoid duplicate findings, and maintain single source of truth.
- [Agent Activation Invocation](https://banes-lab.com/records/algo/agent-activation-invocation.md): For every agent activation, construct a spawn invocation with agent type, prompt, description, model, and context; use forked context for parallel agents and normal context for sequential agents.
- [Parallel Batch Execution](https://banes-lab.com/records/algo/parallel-batch-execution.md): Accumulate adjacent parallel agents into a batch, invoke all spawns in a single concurrent batch, wait for all completions, and merge findings through shared documents.
- [Sequential Agent Execution](https://banes-lab.com/records/algo/sequential-agent-execution.md): Execute state-changing or validation agents one at a time in normal context, pass handoff context from the previous phase, wait for completion signal, then activate the next agent.
- [Four-Dimensional Agent Graph](https://banes-lab.com/records/algo/four-dimensional-agent-graph.md): For each agent, build sequential dependencies, lateral parallel peers, diagonal artifact/data dependencies, and propagation effects including superseded state, propagated contracts, and breakage risks.
- [Handoff Signal](https://banes-lab.com/records/algo/handoff-signal.md): Require every agent to end with a structured handoff signal containing completed agent, phase status, artifact location, next agent, execution mode, context used, findings, files, validation status, parallel results, 4D graph, and orchestrator action.
- [Orchestrator Action](https://banes-lab.com/records/algo/orchestrator-action.md): Interpret handoff action as activate next agent, pause for the developer, or complete workflow; automatically continue unless a critical decision or repeated recovery failure requires the developer's involvement.
- [Workflow Coordination Sequence](https://banes-lab.com/records/algo/workflow-coordination-sequence.md): Pre-create documents, execute parallel batches where eligible, execute sequential agents one at a time, update checkpoints, validate single source of truth, and mark workflow completion.
- [Workflow Recovery Loop](https://banes-lab.com/records/algo/workflow-recovery-loop.md): If an agent fails validation, read shared documents, identify unmet expectation, research the error pattern, relaunch the agent with adapted context, and pause for the developer after bounded recovery attempts.
- [Checklist Integration](https://banes-lab.com/records/algo/checklist-integration.md): Create a workflow progress checklist in the artifact workspace, include execution model metadata, phases in linear dependency order with severity metadata, task checkboxes, progress bars, success criteria, and agent-edit coordination rules.
- [Phase Documentation Template](https://banes-lab.com/records/algo/phase-documentation-template.md): For each agent phase, render phase header, execution mode, context mode, methodology, 4D graph, artifact flow, focus areas, deliverable, and agent activation example.
- [Workflow Principles Mapping](https://banes-lab.com/records/algo/workflow-principles-mapping.md): Derive workflow principles from the agent sequence, first agent grounding role, final agent quality guarantee, intermediate agent contributions, automation, context awareness, recovery, scalability, hybrid execution, context forking, and workspace integration.
- [Capability Invocation Protocol](https://banes-lab.com/records/algo/capability-invocation-protocol.md): Define specialized capability invocation patterns, classify which capabilities run forked or normal, include examples, and allow workflows to call capabilities for research, validation, calculation, and specialized subroutines.
- [Template Assembly](https://banes-lab.com/records/algo/template-assembly.md): Assemble frontmatter, workspace config, overview, phase documentation, checklist requirements, orchestration protocol, handoff format, coordination sequence, capability integration, workflow principles, and domain notes into one workflow artifact.
- [First-Time Initiation](https://banes-lab.com/records/algo/first-time-initiation.md): After generating the workflow template, present a summary and request explicit first-run action: execute workflow, edit template, or cancel.
- [Workflow Validation Gate](https://banes-lab.com/records/algo/workflow-validation-gate.md): Validate DSL compliance, workspace discovery, dynamic agent count, execution classification, context configuration, 4D graphs, handoff protocol, hybrid execution, checklist integration, capability integration, template assembly, dynamic paths, uppercase filenames, and agent-oriented content.
- [Workflow Creation Kernel](https://banes-lab.com/records/algo/workflow-creation-kernel.md): Load DSL and orchestration references, configure file recovery and context forking, discover workspace zones and agents, select shared documents, define dynamic agent sequence, classify execution mode, build 4D graphs, define handoff protocol, assemble coordination sequence, integrate checklist and capabilities, generate phase documentation, assemble template, and validate success criteria.
- [<Workflow Orchestration Concern>](https://banes-lab.com/records/algo/workflow-orchestration-concern.md): <Load orchestration grammar> → <Discover workspace config> → <Select shared documents> → <Define agents> → <Classify parallel vs sequential> → <Fork discovery contexts> → <Serialize action phases> → <Build 4D graphs> → <Create handoff protocol> → <Coordinate shared-document refinement> → <Integrate checklist and capabilities> → <Validate template>
- [Anti-Pattern Inversion](https://banes-lab.com/records/algo/anti-pattern-inversion.md): Take any desired architectural principle, invert its invariants, identify the recurring violation shape, model its propagation mechanism, define detection signals, and derive its remediation inverse.
- [Architecture Smell Record](https://banes-lab.com/records/algo/architecture-smell-record.md): Model every smell as a recurring degeneration path with trigger conditions, enabling conditions, detection signals, damage model, remediation inverse, and prevention gate.
- [Violation Detection](https://banes-lab.com/records/algo/violation-detection.md): For each concept applicable to the current scope, execute its detected_by signals against the implementation, map observations to violated_by patterns, and produce evidence-backed violation records.
- [Canonical Semantics](https://banes-lab.com/records/algo/canonical-semantics.md): Detect duplicated or conflicting models, schemas, terms, rules, and data definitions; select or create canonical authority; normalize variants; enforce single source of truth.
- [Domain Boundary Governance](https://banes-lab.com/records/algo/domain-boundary-governance.md): Identify bounded contexts, ubiquitous language, context relationships, domain model ownership, anti-corruption layers, and explicit boundary rules.
- [Causality and Ordering](https://banes-lab.com/records/algo/causality-and-ordering.md): Model distributed events as a dependency graph, attach causation metadata, validate happens-before relationships, and use sequence numbers, Lamport clocks, or vector clocks when physical timestamps are insufficient.
- [Performance and Scalability](https://banes-lab.com/records/algo/performance-and-scalability.md): Establish workload model, profile runtime behavior, benchmark repeatably, identify bottlenecks, analyze time and space complexity, select scaling strategy, optimize only measured bottlenecks, and enforce SLO gates.
- [Security Governance](https://banes-lab.com/records/algo/security-governance.md): Threat-model the system, reduce attack surface, authenticate identity, authorize actions, validate input, encode output, encrypt data, manage secrets, enforce policy as code, and continuously audit compliance.
- [Metaprogramming Safety](https://banes-lab.com/records/algo/metaprogramming-safety.md): Treat code as data only through schemas, manifests, DSL grammars, reflection contracts, compile-time checks, runtime guards, and generated artifact validation.
- [Model Architecture Governance](https://banes-lab.com/records/algo/model-lifecycle-governance.md): Register models, prompts, datasets, embeddings, retrieval sources, knowledge graphs, inference contracts, evaluation suites, explainability traces, safety gates, and monitoring policies.
- [Canonical Data](https://banes-lab.com/records/algo/canonical-data.md): Normalize incoming data into a canonical schema, validate types and semantics, preserve one source of truth, and translate at system boundaries only.
- [Self-Description Manifest](https://banes-lab.com/records/algo/self-description-manifest.md): Require every component to declare identity, capabilities, contracts, dependencies, configuration, health model, and version metadata in a machine-readable manifest.
- [Causality Ordering](https://banes-lab.com/records/algo/causality-ordering.md): Model events as a dependency graph, assign causal metadata, preserve happens-before relationships, and reject or compensate for invalid ordering.
- [Security Policy](https://banes-lab.com/records/algo/security-policy.md): Threat-model the system, reduce attack surface, authenticate identity, authorize actions, validate input, encode output, encrypt data, protect secrets, and enforce policy continuously.
- [Declarative Metaprogramming](https://banes-lab.com/records/algo/declarative-metaprogramming.md): Represent behavior as data, validate the model or DSL, compile or interpret it into runtime behavior, and restrict reflection or code generation behind safety contracts.
- [Model Lifecycle Governance](https://banes-lab.com/records/algo/ai-model-governance.md): Register models and datasets, version prompts and artifacts, evaluate behavior against benchmarks, validate safety constraints, trace inference inputs, and monitor drift after deployment.
- [RAG Knowledge Boundary](https://banes-lab.com/records/algo/rag-knowledge-boundary.md): Retrieve knowledge from indexed sources, validate relevance and freshness, ground generation in retrieved evidence, and distinguish known, inferred, and unsupported output.
- [Universal Architectural Concern Template](https://banes-lab.com/records/algo/universal-architectural-concern-template.md): For any architectural concern, define its intent, boundary, contract, invariants, allowed variation, forbidden leakage, validation strategy, observability model, and evolution policy.
- [Runtime-Agnostic Adapter Boundary](https://banes-lab.com/records/algo/runtime-agnostic-adapter-boundary.md): Express all workflow operations as semantic verbs, delegate runtime-specific mechanics to adapters, and prohibit repository-, shell-, framework-, model-, or path-specific logic from entering the core contract.
- [Research Guidance](https://banes-lab.com/records/algo/research-guidance.md): Query available external or local guidance, extract best practices and anti-patterns, score research quality, and disclose reduced confidence when current external research is unavailable.
- [Four-Dimensional Phase Graph](https://banes-lab.com/records/algo/four-dimensional-phase-graph.md): For every phase, model the sequential dependency axis, the lateral independent-peer axis, the diagonal shared-data axis, and the propagation axis carrying superseded state, propagated contracts, and what breaks if the edge is omitted.
- [Validation Stage](https://banes-lab.com/records/algo/validation-stage.md): Judge the generated reasoning — not the future implementation — against evidence and semantic policy: run the governance validation suites, verify every material claim by evidence, and apply the semantic-debt rubric before rendering.
- [Evidence-Based Claim Verification](https://banes-lab.com/records/algo/evidence-based-claim-verification.md): Extract every material claim from the generated records, weigh supporting against contradicting evidence in the inventory, and classify each as supported, contradicted, not-applicable, or unsupported — recording the searched scope for a zero-result claim.
- [Cross-Stage Invariants](https://banes-lab.com/records/algo/cross-stage-invariants.md): Bind every stage to the always-and-never rules of the generator — resolve authority and directionality before planning, discover before assuming, activate principles with a bound validator, order by dependency with severity as routing metadata, read only the prior stage's output contract through one evidence-bearing handoff, represent all four graph axes, require evidence for every claim, separate generation gates from future execution gates, repair from the earliest invalid stage, and render deterministically.
- [<Checklist Governance Concern>](https://banes-lab.com/records/algo/checklist-governance-concern.md): <Orient on authority + evidence> → <Plan principles + protocols + dependency graph> → <Compile atomic tasks + ripple chains> → <Validate reasoning by evidence + semantic policy> → <Repair from the earliest invalid stage> → <Render one bounded terminal>
- [Advanced Tool Escalation](https://banes-lab.com/records/algo/advanced-tool-escalation.md): When direct tools cannot answer a verification question, synthesize a specialized analyzer, execute it against the target, parse its output, and integrate the result as evidence.
- [Type-Migration Centralization](https://banes-lab.com/records/algo/type-migration-centralization.md): To move an existing purpose-class or page-scoped stylesheet into this model, treat every appearance-bearing class rule as a migration unit, derive the element type it decorates, register a custom `data-el` type when no tag or variant expresses it, move its look, size, and spacing into one global rule keyed by that type, demote any residual position into components and any residual layout into app, and verify that zero appearance rule keyed by a purpose-class remains outside the globals layer.
- [Profile Compose](https://banes-lab.com/records/algo/profile-compose.md): Specializing the RAG Knowledge Boundary, read the durable profile document (or an empty one), strip volatile metadata, and project a compact view injected into reasoning as retrieved memory rather than ground truth.
- [Delta Capture](https://banes-lab.com/records/algo/delta-capture.md): Specializing the RAG Knowledge Boundary, accept a structured knowledge delta the reasoning turn self-reports inline, admit only facts verified within that same turn, and reject speculation without a separate extraction pass.
- [Living Profile Kernel](https://banes-lab.com/records/algo/living-profile-kernel.md): Compose the accumulated knowledge onto the frozen baseline into a seed, run the reasoning turn, capture the inline verified delta, merge it idempotently under an injected clock, version only real change, and persist against the fork — looping per turn.
- [PAG Handoff Gate](https://banes-lab.com/records/algo/pag-validation-gate.md): Close every node with a handoff gate of three-to-five checks, each a claim about the output with the evidence that settles it and the population it was measured over, a refusal condition named before any irreversible write, the standing of the read set beside the verdict, and a result line that routes pass to the next node, each failure to the earliest node that owns its repair, and unknown to blocked.
- [PAG Invariant Record](https://banes-lab.com/records/algo/pag-constraint-boundary.md): State every behavioural invariant as a record with four slots: the property in a form that could be false, the set it quantifies over, the parties it binds, and the objector, the check that would disagree if the property stopped holding or none as declared debt, so an unwatched invariant is visible rather than assumed.
- [PAG Ambiguity Reduction](https://banes-lab.com/records/algo/pag-ambiguity-reduction.md): Replace interpretive prose with explicit structured tokens so the model completes recognized patterns, while accepting that output stays probabilistic: the grammar reduces input ambiguity, it does not constrain output tokens or guarantee determinism.
- [PAG Authoring Kernel](https://banes-lab.com/records/algo/pag-authoring-kernel.md): Compose a PAG document as frontmatter and typed declaration, then a meta block with jurisdiction, then nodes each headed by its layer, axis, math type and yields, contracted to read the prior node's output, and closed by a handoff gate with evidence, population, refusal and the three verdicts, drawing directives from the semantic operations, control flow and structure declarations, and closing with invariant records and a report.
- [PAG Well-Formedness Validation](https://banes-lab.com/records/algo/pag-well-formedness-validation.md): Scan a PAG document for structural and epistemic defects, each named for the shape it catches — a retired unit head, a node without a gate, a check without evidence, a gate without a population or with an empty one, an unknown left unrouted, a write without a refusal, an artifact without freshness, a node declared twice, an input naming no source, an invariant missing its set, parties or objector, a bare invariant block, a lowercase keyword, a bare FOR, a missing colon, a vague condition — and emit a token-based, regex-free verdict.
- [<PAG Instruction Concern>](https://banes-lab.com/records/algo/pag-instruction-concern.md): <Declare document type and jurisdiction> → <Draw grounded uppercase directives> → <Decompose into headed, contracted, gated nodes> → <Bind semantic operations and structure declarations> → <Bound with invariant records> → <Validate well-formedness against the failure taxonomy>
- [Quality Governance Loop](https://banes-lab.com/records/algo/quality-governance-loop.md): On each update (an apply), normalize the proposal, resolve a CheckPlan, verify it, and on failure fix-then-reverify the downstream set until the verdict is clean or the pass bound is reached, then escalate.
- [Canonical Config Resolution](https://banes-lab.com/records/algo/canonical-config-resolution.md): Expand one canonical config across selected profiles and native rules into per-tool config, resolve exactly one owner per contested surface, detect the four conflict kinds fail-closed, route loosening or self-certification to the developer's gate, and emit a resolved plan.
- [Custom-Rule Derivation](https://banes-lab.com/records/algo/custom-rule-derivation.md): Derive a candidate rule from repeated evidence through anti-pattern classification, boundary-principle evaluation, and candidate selection, migrate with rollback, verify elimination, and admit a tightening rule while routing a self-certifying or loosening rule through the developer's gate.
- [Machine Verdict Derivation](https://banes-lab.com/records/algo/machine-verdict-derivation.md): Derive the verify-stage verdict from the toolchain's machine output — exit codes and parsed findings — never from the model's reading, so a clean verdict is an observed machine fact.
- [Vocabulary Admission Gate](https://banes-lab.com/records/algo/vocabulary-admission-gate.md): Hold every proposed word against the declared vocabularies and the rejection table, resolve and name the declared word that covers it where one does, refuse a word that names a process, an adjective, a grouping label, or a measurement, sort a surviving word by the is-a test into concerns or subjects, and admit it only by developer-approved registry edit.