agent-dashboardlisted

# Agent Dashboard Skill ## Purpose Generate a human-readable local HTML dashboard from `.agent/` coordination state. This is for people to understand progress quickly; machine coordination remains in registry, locks, artifacts, handoffs, and workflows. ## Command ```bash node .agent/skills/agent-dashboard/scripts/generate.js ``` Default output: ```text .agent/metrics/agent-dashboard.html ``` Optional output: ```bash node .agent/skills/agent-dashboard/scripts/generate.js --out .agent/metrics/my-dashboard.html ``` Serve with live updates: ```bash node .agent/skills/agent-dashboard/scripts/serve.js --port 8787 --interval-ms 3000 ``` The server regenerates the HTML periodically and uses browser auto-refresh over SSE. Task cards open a read-only Markdown preview drawer. Related task records, proposals, source refs, artifacts, and gate evidence can be opened without leaving the dashboard. The preview drawer reads documents through: ```text GET /api/preview?path=<project-relative-path> ``` Only `.agent/**`, `docs/**`, and root `README.md`, `AGENTS.md`, `CLAUDE.md`, or `GEMINI.md` files with Markdown, JSON, or text extensions are allowed. Absolute paths, traversal, escaped symlinks, directories, and files over 1 MiB are rejected. ## Data Sources Read these when available: - `.agent/plans/task-progress.md` - `.agent/tasks/*.json` task relations, source refs, artifacts, and gate evidence - `.agent/plans/proposals/**/*.md` files that explicitly reference a task ID - `.a