BHIL-HELM

Heuristic Engine for Layered Mockups. Turn client data into eye-catching, functional, Claude Code friendly interactive dashboards styled in science-fiction design languages. Evidence-tiered. IP-clean. Regenerable.

Read ABOUT.md for the full story of the framework, or open the visual reference pack that the catalog was distilled from.

Fictional user interfaces are the most effective communication design under time compression ever produced: a film screen has one second to read as complex and clear. HELM keeps that discipline and strips the tropes that make screen UIs anti-usable in real products. The result is a dashboard that serves a decision, looks like a command surface, and can be rebuilt by the client's own Claude Code without the lab in the loop.

The starter build. Tactical HUD language, wall-display density, SYNTHETIC badge, evidence labels on the STATED target and the INFERENCE metric, shape-coded status rail, decorative ambient stream. Rendered without the CDN to show the inline fallback.

Sample dashboard library

These ten panels are BHIL concept renders: launch imagery that shows the target fidelity HELM aims at, not builds produced by the pipeline. Each is annotated with the data shape it represents, the matching rule that would fire, and the decision question a HELM build of it would have to answer. Any brand marks, platform names, and faces inside the renders are illustrative placeholders. Full-resolution PNGs are in assets/samples/.

Audience Funnel and SEO Intelligence Unify audience behavior and search performance into one growth read. Data shape: status-grid · Rule M01 binds command-console Decision: Which content cluster gets the next month of effort?	Social Network and Influence Graph Map communities, bridges, and cascade paths across a 64K-node network. Data shape: relationship-graph · Rule M04 binds gestural-holographic Decision: Which bridge nodes move a narrative between communities?
Due Diligence and Risk Intelligence Target overview, risk heat map, governance, red flags, data-room progress. Data shape: evidence-board · Rule M10 binds tabletop-motion Decision: Does the deal proceed to the next review stage?	AI Research Swarm Operations Center Agent roles, workflow stage, evidence tiers, confidence, and a decision gate. Data shape: status-grid · Rule M01 binds command-console Decision: Approve, request more data, or escalate the recommendation?
Influencer Analysis and Narrative Map Reach, influence tiers, narrative clusters, amplification paths, sentiment. Data shape: relationship-graph · Rule M04 binds gestural-holographic Decision: Which narrative is accelerating and who is amplifying it?	Competitive Intelligence War Room Competitor snapshot, market motion, feature benchmark, pricing shift, signal radar. Data shape: multi-unit · Rule M06 binds hard-realism-tactical Decision: Which competitor move requires a response this week?
Partner Ecosystem Intelligence Deal flow, relationship strength, dependency map, ecosystem opportunities. Data shape: evidence-board · Rule M10 binds tabletop-motion Decision: Which partnership opportunity gets pursued this quarter?	OSINT Command Center Signal intake, source credibility, geo activity, entity resolution, timeline. Data shape: threat-network · Rule M05 binds sonar-surveillance Decision: Which priority signal needs analyst attention now?
Counterintelligence Network Monitor Threat posture, network graph, alias matrix, travel routes, exposure. Data shape: threat-network · Rule M05 binds sonar-surveillance Decision: Which target network carries the highest risk score today?	Talent Intelligence Command Surface Hiring velocity, skill clusters, role demand, compensation bands, geography. Data shape: status-grid · Rule M01 binds command-console Decision: Where is the labor market moving against the hiring plan?

What is in the box

Quick start

git clone https://github.com/PolymathWizard/BHIL-HELM-Sci-Fi-Visualizer
cd BHIL-HELM-Sci-Fi-Visualizer
python3 tools/validate.py                      # PASS: 15 languages, 12 tropes, 5 tiers, 11 rules, 14 components
python3 tools/contrast.py tactical-hud         # WCAG matrix for one language
open examples/helm-starter/dashboard.html      # no build step, no server

Then in Claude Code at the repo root:

/helm brief

The skill confirms six intake values, creates engagements/<REF>/, and runs profiler, stylist, builder, and QA in order, asking after each stage whether to continue or refine.

How it works

DATA ──► SP-1 Data Register ──► SP-2 Decision Map ──► SP-3 Language + Narrative Tax
                                                              │
                              SP-4 Zones ──► SP-5 Components ──► SP-6 Artifact + scaffold
                                                                        │
                                                    SP-7 Motion ──► SP-8 QA ──► SP-10 SHIP / HOLD / REWORK

Profile before styling. One language per screen. Every number carries an evidence class (VERIFIED / CORROBORATED / UNCORROBORATED / INFERENCE / STATED) and the lower three render their label on screen. Ambient content is marked decorative. The builder never gates its own output.

The catalog in one table

Full lineage cards, tokens, and motion grammar: docs/catalog.

What HELM refuses to do

1. Reproduce a protected work. Lineage is a study reference. No franchise name, logo, glyph system, licensed font, or copied frame enters a deliverable. The test suite enforces it.
2. Fabricate data to fill a zone. Empty is honest. Synthetic is labeled. Ambient is decorative.
3. Ship a dashboard the client cannot regenerate. Every build carries CLAUDE.md, a skill, and REGENERATE.md.

Commercial tiers

Snapshot (one screen, two-page readout, the door-opener) · Core Build (full brief, artifact, scaffold, QA) · Enterprise (multi-screen, multi-language token sets, design-system handoff) · Retainer (rebind, restyle, add-panel cycles). Details in docs/guide/commercial.md.

Repository layout

BHIL-HELM-Sci-Fi-Visualizer/
├── CLAUDE.md                   standing brief for any Claude Code session
├── prompts/                    00-master + sp-01 .. sp-10
├── data/canonical/             catalog, tropes, evidence tiers, matching, components, lineage register
├── data/schemas/               JSON Schema 2020-12 with count laws
├── tokens/                     generated CSS token sets (15)
├── templates/                  artifact, CLAUDE.md, REGENERATE.md, client SKILL.md
├── examples/helm-starter/      complete starter build
├── assets/hero/                launch imagery (BHIL originals, full resolution)
├── assets/samples/             ten sample dashboard concept renders (full resolution)
├── research/reference-pack/    study board: 43 sourced visuals, coverage ledger, catalog PDF/HTML
├── .claude/                    agents, commands, skills, settings
├── tools/                      validators and builders (stdlib only)
├── tests/                      regression tests naming specific bugs
├── docs/                       MkDocs site (derived pages generated)
├── launch/                     GitHub description options and LinkedIn posts
├── ABOUT.md                    what HELM is, why, and how it runs
├── DISCLAIMER.md               third-party material, intended use, no warranty
└── engagements/                gitignored client work (.gitkeep only)

Disclaimer

All film, television, and game titles, studio and designer names, and the study images in research/reference-pack/ are the property of their respective license holders and appear here for design study and lineage attribution only. This repository is solely for development, research, and innovation use. See DISCLAIMER.md.

Conventions

Canonical JSON generates every derived artifact; CI fails on drift. Dual licensed: MIT for code, schemas, and tokens; CC BY 4.0 for prose. Conventional commits scoped helm:. No em dashes in shipped prose. See CONTRIBUTING.md.

Related BHIL frameworks

CODEX (agent scaffolding, upstream) · CDPA (data architecture, downstream) · QUADRA, CADRE, FACET, LOCUS (the BHIL framework family) · BHIL-AI-First-Development-Toolkit

HELM. Barry Hurd Intelligence Lab. Human-Directed. AI-Enabled. Commercially Tested.