Audit produced by running the §2.6 mandatory stub-indicator regex
(scripts/stub-indicator-scan.sh) plus the broader CLAUDE.md scan
((would|will) (use|call|invoke|fetch|implement|need|require),
in production[,:], in a (real|full) (implementation|integration),
simulate, not yet implemented, TODO:, FIXME:, HACK:, placeholder,
hardcoded, hard-coded, \bfake\b) over apps/lilith/ and libs/lilith/ on
2026-05-04.
Scope. Production source (*.ts, *.tsx) under apps/lilith/ (84 service
apps plus web, mobile, desktop, bff, cli, contracts, shared) and
libs/lilith/ (22 packages including service-lib, fastify-core, vectordb,
database, cache, etc.). Excluded: node_modules/, dist/, *.test.ts,
*.spec.ts, *.stories.tsx, locales/ directories (i18n strings), and lines
already marked // stub:legitimate <reason>.
The strict pre-commit scan
(scripts/stub-indicator-scan.sh apps/lilith/ libs/lilith/) returns exit 0
for both --mode=stubs and --mode=all (which adds the silent-Math.random()
check) — the scanner's allow-list has already absorbed the historical Lilith
hits.
Counts#
| Bucket | Files | Line hits |
|---|---|---|
Strict scan (stub-indicator-scan.sh defaults) |
0 | 0 |
Strict scan (--mode=all, includes random) |
0 | 0 |
| Broader CLAUDE.md grep (production source) | 306 | 978 |
Density distribution of the broader-grep hits across files:
| Density (hits per file) | File count |
|---|---|
| ≥10 | 4 |
| 5–9 | 42 |
| 2–4 | 109 |
| 1 | 151 |
Explicit V1-P2-2501..2503 targets#
These are called out explicitly in OSHUN_V1_TODOS_P2.md and have dedicated tasks; the per-file checklist below also includes them so V1-P2-2504 can walk the full list in one pass.
- V1-P2-2501 —
apps/lilith/svc-ai/src/app.impl.ts:2191—score: Math.random() * 0.3 + 0.7. Real fake AI score; replace with actual model output. Adjacent random sites at lines 1357, 1465, 1470, 1473 (response-time and confidence fakes) should be wired in the same pass. - V1-P2-2502 —
apps/lilith/mobile/src/services/knowledge/ConceptExplorationService.ts:1232—relevance_score: Math.random() * 0.5 + 0.5. Real fake relevance; replace with embedding cosine similarity against the concept vector. - V1-P2-2503 —
apps/lilith/svc-metaverse/Math.random uses (236 total, across 8 production files plus tests). Bucketed under V1-P2-0353's allow-list policy:- Allow-list as procedurally generative: CosmicSpaces, UnderwaterConsciousness, ProceduralWorldGenerator, AstralProjection, EnergyFieldVisualization, WeatherSystem, AssetLibrary, TerrainGenerator. These intentionally synthesize procedural worlds and energy field perturbations — Math.random is the API surface, not a stub.
- Replace: any Math.random hits found in service routing, ID generation,
or metric reporting elsewhere in svc-metaverse should use
randomUUID()/randomInt()fromnode:crypto.
High-density files (≥10 hits) — sweep first#
-
apps/lilith/svc-multi-region-resilience/src/app.ts— 37 hits. The service is an explicit in-process simulator (lines 34-45 ship a production guard refusing to start withNODE_ENV=productionunlessMULTI_REGION_BACKEND=realis also set). Most hits are legitimate — the simulator vocabulary is the API surface — but thesimulate*Failover()factory functions and per-region health score generators should carry explicit// stub:legitimate simulator-by-designannotations so the broader regex stops flagging them. -
apps/lilith/web/src/features/publishing/PresentationGeneration.tsx— 35 hits. Mostly UI copy ("simulate the export...") for meditation-author tooling. Real stub if the export pipeline is meant to actually produce presentations; investigate the generate/export flow before classifying. -
apps/lilith/web/src/accessibility/visual/color-blind.ts— 11 hits. Per-modesimulate*Vision()color transformations are the core API of a colorblind-simulator module — legitimate domain, annotate. -
libs/lilith/service-lib/comfyui/workflow-builder.ts— 11 hits. ComfyUI image-generation workflow construction. Inspect: many hits are likely "TODO: support workflow X" comments; classify individually.
Medium-density files (5–9 hits) — 42 files#
Top entries (full list in /tmp/lilith-counts.txt from the audit run):
-
apps/lilith/svc-ai/src/app.impl.ts(covers V1-P2-2501) -
apps/lilith/web/src/features/publishing/AcademicFormatting.tsx -
apps/lilith/web/src/components/SeamlessAuthManager.tsx -
apps/lilith/mobile/src/components/meditation/JournalEntry.tsx -
apps/lilith/web/src/features/workspace/NoteEditor.tsx -
apps/lilith/web/src/features/references/CitationImport.tsx -
apps/lilith/web/src/features/community/ScholarProfiles.tsx -
apps/lilith/svc-content/src/meditations/meditation-repository.ts -
apps/lilith/svc-ai/src/image-moderation/patterns.ts -
apps/lilith/mobile/src/screens/AuthScreen.tsx -
apps/lilith/mobile/src/components/molecules/Select.tsx(UI inputplaceholderattribute — legitimate) -
apps/lilith/desktop/src/components/CapturePanel.tsx -
apps/lilith/web/src/features/zettelkasten/BlockEditor.tsx -
apps/lilith/web/src/features/community/QASystem.tsx -
apps/lilith/web/src/features/community/DiscussionForums.tsx -
apps/lilith/web/src/features/collaboration/PeerReview.tsx -
apps/lilith/svc-voice-pipeline/src/simulation-routes.ts(filename announces the simulation purpose — legitimate) -
apps/lilith/contracts/scripts/deployment-security.ts -
apps/lilith/web/src/features/workspace/Bookmarks.tsx -
apps/lilith/web/src/features/collaboration/ResearchGroups.tsx -
apps/lilith/web/src/components/ProgressTracker.tsx -
apps/lilith/svc-transaction-manager/src/app.ts -
apps/lilith/svc-operational-excellence/src/health-monitor.ts -
apps/lilith/svc-notification/src/dispatch/providers/mock.ts(filename announces the mock purpose — legitimate) -
apps/lilith/svc-media/src/asset-repository.ts -
apps/lilith/svc-ai/src/social-community/stream-transport.ts -
apps/lilith/contracts/scripts/deploy-multichain.ts - (16 additional files at 5 hits each — see
/tmp/lilith-counts.txt)
Low-density files (2–4 hits) — 109 files#
Spread across web features, mobile services, BFF utilities, and many backend
services. The triage cost is roughly proportional to the per- file hit count;
address the high-density files first and the long tail becomes mostly UI
placeholder attributes plus occasional "in production..." comments to rewrite.
Single-hit files — 151 files#
Mostly UI input placeholder attributes (HTML attribute, not a stub marker) and
a long tail of one-off "in production..." comments. Many of these will resolve
with a single // stub:legitimate annotation or a one-line comment rewrite.
Per-directory hot-spots#
Hits per leaf directory (top 10):
| Directory | Files |
|---|---|
apps/lilith/web/src/components |
12 |
apps/lilith/svc-metaverse/src/ai |
7 |
apps/lilith/web/src/features/workspace |
6 |
apps/lilith/web/src/features/collaboration |
6 |
apps/lilith/web/src/components/ui |
6 |
apps/lilith/contracts/scripts |
6 |
apps/lilith/web/src/features/publishing |
5 |
apps/lilith/svc-ai/src/multi-platform-delivery |
5 |
apps/lilith/svc-ai/src |
5 |
apps/lilith/web/src/features/zettelkasten |
4 |
Summary of action types#
| Class | Approximate count |
|---|---|
| Real stubs to wire | ~60–80 |
| Legitimate domain — annotate | ~150 |
| Legitimate UI input attributes | ~80 |
| Already-correct, comment-only | ~50 |
V1-P2-2504 will iterate this list, classifying each entry on first read and
either wiring the implementation, adding a // stub:legitimate <reason>
annotation, or rewriting the comment to honestly describe the chosen
implementation. The strict scanner's continued green status (exit 0) is the
source-of-truth signal for completion per CLAUDE.md project policy.