Implement Phase 1: single-agent scoreboard dashboard

[rdwj]

Summary

Implement the Phase 1 single-agent dashboard described in docs/architecture.md.

Deliverables

• index.html — single-file dashboard with inline CSS and vanilla JS
• Panels: health indicator, agent-info header, trust gauge, maturation stage badge + progress bar, skill list, work-item queue depth bars
• Polls one agent URL on a configurable interval (default 10s)
• AGENT_URL configurable via query param (?agent=http://...) or env var
• Graceful degradation: panels show "N/A" when endpoints return 404
• Containerfile — UBI nginx serving the static file
• chart/ — Helm chart (Deployment, Service, Route, ConfigMap for AGENT_URL)
• Makefile — make build, make deploy PROJECT=<ns>
• Basic smoke test: container starts, / returns 200, /healthz returns 200

Endpoints to poll

Endpoint	Panel	Required
GET /healthz	Connection dot	Yes
GET /v1/agent-info	Header (name, version)	Yes
GET /v1/agent/trust	Trust level + score	No
GET /v1/agent/maturation	Stage badge + progress	No
GET /v1/agent/skills	Skill count + list	No
GET /v1/work-items/stats	Queue depth bars	No

Design constraints

• No npm, no build step, no CDN dependencies
• Red Hat UBI base image
• Must work behind OpenShift OAuth proxy (no built-in auth)
• Keep index.html under 500 lines

Future phases (not this issue)

• Phase 2: Multi-agent support (AGENT_URLS, gateway auto-discovery)
• Phase 3: History sparklines (inline SVG, no chart library)

See docs/architecture.md for the full design.