# Infinite Sandbox 0.0.13a_03 — certification-readiness evidence index

## Status and claim boundary

Infinite Sandbox 0.0.13a_03 is an experimental educational and research
prototype. It is not certified, accredited, professionally validated, or
approved for real-world engineering decisions. Automated results demonstrate
only the coded behavior covered by their tests.

## Release controls

- Runtime and HUD version: `0.0.13a_03`
- npm package version: `0.0.13-alpha.3`
- Engineering model: `infinite-sandbox-vvuq-kernel-2`
- Fluid model: `bounded-shallow-fluid-3`
- Planet model: `planet-local-tangent-1`
- Mentor model: `grounded-structured-mentor-2`
- Complete automated gate: `npm run verify:release`

The release regression checks pointer-based left/right editing, compatibility
release handling, one-cell placement retry, removal of the N shortcut,
adaptive-CFL fluid behavior, closed-grid volume balance, water texture
assignment, smooth per-vertex alpha, and normalized fluid surface normals. The
3×3 depth smoothing is explicitly a render filter and does not alter the fluid
state used by the solver. Production checks require the 0.0.13a_03 identifier
and this evidence set to ship in the bundle.

## Evidence groups

| Area | Included evidence | External work still required |
| --- | --- | --- |
| Release behavior | `RELEASE_0.0.13a_03.md`, input/source tests and production smoke test | Physical-device and assistive-technology review |
| Fluid credibility | `ENGINEERING_VVUQ.md`, adaptive-CFL and conservation tests, SI runtime diagnostics | Grid/time convergence, calibrated inputs, experimental and field validation, independent hydraulic review |
| Structural credibility | SI runtime console, solver analytic tests, uncertainty kernel and `level.dat` records | Material calibration, application validation, qualified independent review |
| Educational design | `EDUCATIONAL_ASSURANCE.md`, `RESEARCH_BASIS.md`, adaptive-policy and learner-journey tests | Educator review, safeguarded studies, validated outcome measures |
| Responsible AI | `AI_BIAS_AND_OVERSIGHT.md`, local audit trail, override and concern-export controls | Operational reporting, fairness study, human oversight process |
| Learner variability | `LEARNER_VARIABILITY.md`, support/pace controls, responsive hints, touch controls | Practitioner co-design and formal accessibility testing |
| Privacy and security | `PRIVACY.md`, `SECURITY.md`, restrictive CSP and data-minimizing tests | Deployment threat model, operations, legal review, incident process |
| Certification mapping | `CERTIFICATION_READINESS.md`, `FERPA_READINESS.md`, `QUALITATIVE_STUDY_PROTOCOL.md` | Licensed/current requirements, applications, external review, approval |

## Standards-informed organization

The evidence structure is informed by the public ASME VVUQ portfolio,
NASA-STD-7009B and NASA-HDBK-7009B credibility guidance, NIST SI references,
and public USACE shallow-water numerical guidance. Referencing those materials
does not claim conformance. A future professional assessment needs a defined
intended-use domain, licensed requirements where applicable, traceability,
qualified independent reviewers, and validation evidence.

## Interpretation rule

Passing every included test must be reported only as “automated software checks
passed.” It must not be reported as certified, standards-compliant,
engineering-grade, accessible, effective, safe for professional decisions, or
approved by an outside organization.

The authoritative machine-readable inventory is
`certification-evidence-0.0.13a_03.json`.