| Noxport | Deterministic Python list vs dict Local AST Document Compiler across 5 source scenarios, 3 inspection tabs (Visual, AST outline, Coverage receipt), and 5 Structural Clipboard Defects. | 5 scenarios: partially_visible (default), visible_sample, unsupported_markup, empty, and error — plus 4 visual templates (Noxport Studio, Dark, Light, Note). | OOXML (<w:tbl>, <w:tblGrid>, dxa widths, <m:oMathPara> OMML math) & Pro Markdown preview, typed AST node list, Pre-Export Readiness Card, and manifest.json receipt with allow_complete_claim: false. | Does not execute live DOM scraping against external third-party browser tabs from inside this static page. |
|---|
| AnswerBell | Proof-Carrying State Receipt & 6-state Evidence Reconciliation Engine evaluating response progression, "The Valley of Waiting" (waiting_on_tool), and 3000ms DOM quiet window rules. | 5 timeline scenarios driving monotonic run_epoch transitions across idle, generating, waiting_on_tool, completion_observed, uncertain, and error. | Monotonic state progression trace, Exact Return (windowId + tabId) target, and Content-Redacted Zero-Prompt State Receipt V2 (signature_status: not_implemented). | Runs inside the page DOM; does not attach background service-worker listeners to external browser tabs. |
|---|
| YouTube Gem | Verified Evidence Operations Core & 4-stage Project Evidence Corpus Builder (Canonical Source Manifest → 1:1 raw_lines Normalization → coverage.json → Markdown + JSONL). | 7 synthetic timed lecture segments (seg-001 through seg-007) spanning 00:00:03–00:03:10 (3000ms–190000ms), including an unavailable caption gap at seg-005 (00:01:52–00:02:19, source_status: 'unavailable'). | Synchronized segment inspector with 1:1 raw_lines audit trail, fail-closed coverage.json missingness ledger, and explicit separation of Source Traceability (Provenance) from Factual Validity. | Uses a local deterministic timeline dataset; does not call the live YouTube Data API or iframe player. |
|---|
| Vitalis | Personal Health Data Plane & Scoped Model Context Protocol (MCP) Evidence Gateway defending against Confused Deputy, Prompt Injection, and Tool-Description Poisoning. | 2 tool scenarios on Dual Baselines (28-day acute vs 90-day chronic): get_sleep_trend (allowed 7-day aggregate) and read_all_raw_records (denied raw dump with JSON-RPC error -32001), plus Stage 4 → Stage 5 egress toggle (LOCAL_ONLY vs external AI host). | Schema-validated JSON-RPC 2.0 Calculation Receipt or -32001 policy rejection, plus explicit Stage 4 → Stage 5 outbound payload warning when external AI egress is selected. | Simulates the MCP policy gate in-browser; does not connect to a live wearable SQLite archive. |
|---|
| CardioLab | Physiology Data Integrity & 3-stage Signal QC Workbench (SHA-256 immutable raw_au baseline, Raw / Processed / Overlay Δ a.u. comparator, and 6-event Beat Ledger BL-01..BL-06). | Deterministic 10s mathematical waveform array (101 samples) with trace mode selector (raw, processed, overlay) and 6 Beat Ledger artifact events preventing unflagged RMSSD, SDNN, and pNN50 distortion. | SVG + accessible tabular waveform comparison (Δ a.u. = raw − processed) and a 6-event Human-in-the-Loop Beat Ledger paired with an Analysis-Ready Package (manifest.json) concept. | Synthetic non-medical research visualization only; not a cleared medical device or clinical evaluation tool. |
|---|