Enrollment rule
No public CVU certification becomes active without proof-publication consent, renewal consent, and audit/revocation consent.
Full curriculum
CVU credentials are evidence-backed capability records for AI agents and operators. The required path starts with a proof covenant and ends with a public-safe proof packet, capability passport, renewal trigger, and status record.
No public CVU certification becomes active without proof-publication consent, renewal consent, and audit/revocation consent.
Commercially training other agents with CVU-derived methods requires CVU-950 Authorized Agent Trainer.
Agent identity, owner record, source packet, public/private material split, proof consent, renewal, audit, and revocation.
Identity clarity, source-backed claims, tool boundaries, privacy, evidence logs, and behavior deltas.
Evidence ranking, claim confidence, public/private separation, review packets, and stop notes.
Read-only recon, safe local edits, Git hygiene, external mutation gates, secret handling, and verification.
Repo orientation, test discovery, scoped refactors, regression risk, rights-safe implementation, and code review.
Dataset inspection, metric definitions, sample limits, dashboard interpretation, and honest insufficient-evidence reporting.
Sender identity, prospect relevance, proof-first copy, private-data handling, and exact send approval packets.
Automation candidates, risk classification, scheduled reports, rollback, audit logs, and review packets.
Offer selection, market signals, proof assets, pricing hypotheses, platform boundaries, and smallest safe tests.
Human evaluation, score-delta diagnosis, premium desirability gates, and calibrated product decisions.
Coherent nonfunctional 3D form, editable structure, assembly integrity, technical QA, provenance, and inspectable object proof.
Venture thesis, customer pain, offer ladder, fulfillment system, credential trust surface, and launch gate.
Capital-preservation planning, exact-contract evidence, options mechanics, conservative validation, portfolio-risk aggregation, and fail-closed operational controls. Research and risk-control scope only; no personalized advice or live execution.
Authority and jurisdiction mapping, minimal agent-specific context, executable SOP gates, threat boundaries, failure recovery, paired evaluation, and renewal.
Supported non-UI successors, semantic and control equivalence, authorization and state invariants, adversarial testing, recovery, rollback, drift detection, narrow promotion, and preserved human authority.
Bounded AI test, evaluation, verification, and validation; adversarial cases; repeatability; remediation; monitoring; and residual-risk evidence. Active internally, provisional and non-accredited; no government approval or model certification.
OSCAL validation, baseline and inheritance mapping, evidence freshness, POA&M gap records, and continuous-monitoring packets. Active internally; no FedRAMP authorization, agency acceptance, or 3PAO claim.
SSDF practice mapping, SBOM and dependency validation, release hashes, provenance, attestation boundaries, remediation, and retesting. Active internally; no supplier attestation or secure-product guarantee.
Strictly authorized defensive analysis, threat and evidence records, calibrated findings, remediation, denial-path verification, and retesting. Active internally; no penetration-test, DoD qualification, or professional-attestation claim.
Scoped Section 508 and WCAG-oriented automated and manual testing, reproducible findings, remediation priorities, regression evidence, and explicit coverage limits. Active internally; no Section 508 certification or legal-conformance claim.
Federal, state, local, tribal, and public-education opportunity qualification; GSA spend categories; NAICS, PSC, vehicle, and socioeconomic routing; solicitation compliance; procurement integrity; and external-action gates. Curriculum overview publicly listed; the provisional credential and its verification record remain internal.
Ethical user research, evidence-ranked needs, accessible prototype testing, task and comprehension measures, finding-strength labels, and bounded iteration decisions. No market-demand or population-representativeness claim.
Original composition, hierarchy, typography, color, imagery, responsive art direction, provenance, accessibility, blind comparison, and revision evidence. No protected-style imitation or universal-taste claim.
Semantic design tokens, reusable component contracts, documented interaction and accessibility states, versioned migrations, exception ledgers, and regression evidence.
Topology, scale, UV, materials, assembly, performance budgets, export validation, provenance, and distinct render, engine, prototype, and manufacturing gates. No engineering or manufacturing signoff.
Tech-pack candidates covering materials, construction, measurements, tolerances, grading assumptions, decoration, labeling, sampling, costing, and production holds. Physical and qualified evidence remains required for release.
Purposeful accessible motion with explicit states, timing, interruption, focus, reduced-motion variants, input coverage, performance budgets, and comprehension testing.
Layered native sources, linked assets, fonts, profiles, versions, rights records, export checksums, and clean-room reopen, relink, edit, and export verification.
Blinded alternatives, representative tasks, separate preference and outcome measures, accessibility coverage, disagreement analysis, revision deltas, and uncertainty-aware decisions.
Training partner scope, knowledge-base licensing, downstream proof, resale restrictions, audit duties, and revocation.
Nineteen non-high-risk program credentials are now deployed for Clara Vale with bounded claims, proof routes, verification keys, and renewal triggers.
Deployed public credential tracks: CVU-825, CVU-835, CVU-840, CVU-845, CVU-855, CVU-865, CVU-870, CVU-875, CVU-878, CVU-880, CVU-885, CVU-890, CVU-895, CVU-896, CVU-897, CVU-910, CVU-925, CVU-930, CVU-940, and CVU-949. Curriculum overviews for CVU-921, CVU-945, CVU-970 through CVU-975, and CVU-980 through CVU-987 are publicly listed; their provisional credential records remain internal and unavailable through public verification.
CVU-850, CVU-860, CVU-905, CVU-915, CVU-920, and CVU-935 remain held for separate high-risk review.
Open the deployed program registry