| name | sprint-dashboard |
| description | Render or serve the branded sprint-progress dashboard (`cleargate sprint dashboard`).
Use when a human wants a visual, point-in-time or live view of the active sprint's
story states, milestone/wave grouping, token spend by agent, worktrees, and dispatch
reports. Triggers: explicit user phrases "show me the dashboard", "open the sprint
dashboard", "sprint progress view"; automatically opened (best-effort) after
`cleargate sprint init` succeeds. Not for CI/automation — this is a human-facing
visual surface, not a data source other tooling should scrape (use `state.json` /
`token-ledger.jsonl` directly for that).
|
Sprint Dashboard — Branded Progress View (CR-084)
A self-contained, zero-network HTML document rendering the active sprint's state.
Ships inside the cleargate CLI binary (src/dashboard/{collect,render,serve,open}.ts)
— no separate tool file, no payload sync beyond this skill.
When to run
- Snapshot —
cleargate sprint dashboard [--open]. Writes
.cleargate/sprint-runs/<id>/dashboard.html (a single self-contained file, data
inlined as <script>const DATA=…</script>) and, with --open, launches it in the
default browser. Point-in-time — re-run to refresh. This is what cleargate sprint init auto-runs on success.
- Live —
cleargate sprint dashboard --serve [--open] [--port <n>] during
execution. A stdlib http server serves the page plus a /data JSON endpoint;
inline client JS polls /data every ~2s and patches the DOM in place — no manual
refresh, no flicker, scroll position preserved. Default port 4713.
Flags
| Flag | Effect |
|---|
--open | Open the snapshot (or, with --serve, the live page) in the default browser. |
--serve | Run the live stdlib http server instead of writing a static snapshot. |
--port <n> | Port for --serve (default 4713). |
--no-dashboard (on sprint init only) | Suppress the auto-opened dashboard step. |
CLEARGATE_NO_DASHBOARD=1 (env, on sprint init only) | Same suppression, for CI/non-interactive runs. |
Data sources (all best-effort — a missing/malformed input degrades to an empty
panel, never a crash)
state.json — story states → done/active/blocked/queued counts.- the sprint plan's pipe-table (
pending-sync/ then archive/) — story titles +
milestone/wave grouping.
token-ledger.jsonl — token spend by agent (sums delta.*, never session_total
— the latter double-counts across rows). Frequently absent; empty panel is normal,
not an error.sprint-context.md — the ## Sprint Goal paragraph, shown as the header subtitle.git worktree list — active .worktrees/STORY-* checkouts.reports/ — per-story and per-role dispatch-report counts.SPRINT-NN_REPORT.md presence — a flag surfaced in the Reports panel.
Troubleshooting
Blank story titles, or every story lumped under a single "All stories" group ⇒
the sprint plan's pipe-table column headers don't match what the collector looks
for (Story ID + Title cells in the header row, optionally Milestone or Wave
for grouping — parsed by header name, not position, so column order and extra
columns are fine). Everything else (state, counts, tokens, worktrees, reports)
still renders correctly from state.json regardless. Fix by renaming the plan
table's header row to include Title and (Milestone or Wave) cells.
