Jira Sprint Dashboard
Build a focused dashboard that helps an engineering manager, tech lead, or
senior engineer see current Jira work quickly enough to decide what needs
attention. The output is a dashboard, not a prose report and not a generic
health score.
This skill is read-only by default. Do not create, update, transition, assign,
or comment on Jira work items unless the user explicitly asks for a write action
after reviewing the dashboard.
Output Mode
Use the richest dashboard renderer supported by the current environment. The
dashboard content, claims, counts, and source appendix must stay consistent
across renderers; only the presentation changes.
Choose the renderer in this order:
- Cursor Canvas, if running in Cursor with Canvas support.
- Interactive artifact, if the current agent supports HTML, React, or similar
artifact output.
- Static HTML file, if file creation is available and useful.
- Markdown dashboard, if no richer visual renderer is available.
- Structured JSON plus concise summary, only if visual rendering is impossible.
Do not mention that Cursor Canvas is unavailable unless the user specifically
asked for Cursor Canvas. If the user asked for a dashboard generally, use the
best available renderer without apologizing for the environment.
Cursor Canvas Renderer
Use this section only when running in Cursor with Canvas support.
Read ~/.cursor/skills-cursor/canvas/SKILL.md before writing canvas code. If
you need exact exports or prop shapes, read the files in
~/.cursor/skills-cursor/canvas/sdk/.
Canvas constraints:
- Create one
.canvas.tsx file in the Cursor canvases directory.
- Import only from
cursor/canvas. Do not import react, CSSProperties,
JSX, Atlaskit, or other packages.
- Embed Jira data inline in the canvas; do not fetch from the canvas.
- Prefer Canvas primitives such as
Stack, Grid, Card, Stat, Table,
Pill, Callout, UsageBar, BarChart, LineChart, PieChart, and Code
over raw HTML.
- Use
useHostTheme() for custom styles. Do not hardcode hex colors, gradients,
box shadows, ADS variables, unsupported CSS frameworks, or @atlaskit/*.
- Do not publish or share the canvas unless the user asks.
Portable Renderers
Use this section when Cursor Canvas is unavailable.
For an interactive artifact renderer:
- Render the same dashboard model as an interactive artifact.
- Prefer tables, compact charts, stat rows, and collapsible source details.
- Keep interactions lightweight: filtering, expanding details, or switching
chart/table views is fine; do not require live Jira fetching from the artifact.
For static HTML:
- Create a self-contained dashboard file with embedded data.
- Use responsive layout, accessible tables, and simple chart-like visuals when
chart libraries are unavailable.
- Do not fetch Jira data from the HTML file.
For Markdown:
- Preserve the dashboard order.
- Use compact tables for stats, owner load, risks, highest-priority work, and
source appendix.
- Use textual chart substitutes only when they remain honest, such as counts,
percentages, and simple bars.
- Avoid turning the output into a long prose report.
For JSON fallback:
- Return the normalized dashboard model.
- Include a short human-readable summary with the highest-signal risks and the
source scope.
Get The Scope
Do not guess the Jira scope. If the user does not provide a project key, space
key, board, sprint, filter, JQL, work item keys, or Jira URL, stop and ask for
one. A dashboard from a random visible project or guessed team context is worse
than no dashboard.
If the user gives a project or space key but no sprint, board, or filter, start
with the Jira JQL project field and the user's key:
project = "SPACE_KEY" AND sprint in openSprints() ORDER BY Rank ASC
If the open sprint result is empty, stale, or misleading, switch to snapshot
mode and say so in a compact caveat below the top bar:
project = "SPACE_KEY" AND statusCategory != Done ORDER BY priority DESC, updated ASC
project = "SPACE_KEY" AND updated >= -60d ORDER BY updated DESC
Use a 60-day recent movement window by default unless the user asks for another
period.
Not every board supports sprints. Team-managed boards of type simple have no
sprints at all: sprint in openSprints() returns nothing and a board-sprint
lookup fails with "The board does not support sprints". Treat that as a signal to
go straight to snapshot mode rather than an error worth retrying.
Query Jira
Use read-only Jira search. Request only fields needed for the dashboard and
tolerate missing fields.
Useful fields: key, summary, status, statusCategory, assignee,
priority, issuetype, created, updated, resolutiondate, duedate,
parent, issuelinks, labels, components, fixVersions, sprint, and any
available estimate/story point field.
Sprint and story points live in custom fields, which the default response omits.
Jira search defaults to a compact view, so a sprint dashboard built from it will
look sprint-less even when the sprint is active. To get them, either pass
view: "evidence" (or "full"), or request this site's customfield_* IDs
explicitly — the IDs differ per Cloud site. Custom field values come back under
fields.customFields, not as top-level customfield_* keys.
Start with maxResults: 100. For complete sprint, board, or filter dashboards,
paginate until the scope is complete or too large for useful work-item-level
rendering.
Default to one complete paginated scope query. Derive ordinary dashboard signals
locally from the returned work item set instead of issuing separate JQL calls for
each signal.
Derive these locally when the scope query returned the required fields:
- Recently completed from
statusCategory = Done and resolutiondate.
- Aging unfinished from
statusCategory != Done and updated.
- Unowned unfinished from
statusCategory != Done and empty assignee.
- High-priority unfinished from
statusCategory != Done and priority.
- Status or label blockers from
status, statusCategory, and labels.
- Owner load, stale work, due date risk, and planning gaps from the normalized
scope dataset.
Use targeted follow-up queries only when they are needed to support a visible
claim that cannot be derived safely from the scope data, when the scope is too
large for useful local processing, or when the user asks for an audit-style
dashboard with exact evidence per signal.
Rule of thumb:
- Small or medium sprint dashboard: prefer one full paginated scope query, with
at most one targeted blocker-text or dependency-status follow-up when needed.
- Large scope dashboard: use narrower follow-up queries when pulling and
processing the full work-item set would be slow or low-value.
- Evidence-heavy review: multiple focused queries are acceptable when the exact
JQL evidence matters more than minimizing calls.
Examples of targeted follow-up queries, only when justified:
- Recently completed:
<scope> AND statusCategory = Done ORDER BY resolutiondate DESC
- Aging unfinished:
<scope> AND statusCategory != Done AND updated <= -3d ORDER BY updated ASC
- Unowned unfinished:
<scope> AND statusCategory != Done AND assignee is EMPTY ORDER BY priority DESC, updated ASC
- High-priority unfinished:
<scope> AND statusCategory != Done AND priority in (Highest, High) ORDER BY priority DESC, updated ASC
- Blocked signal:
<scope> AND statusCategory != Done AND (status = Blocked OR text ~ "blocked" OR labels in (blocked, blocker)) ORDER BY priority DESC, updated ASC
Do not make negative claims such as "no blockers" or "no dependencies" unless
the source appendix shows the query or returned field coverage that supports the
claim. If only status and labels were checked for blockers, say that no
status/label blockers were found rather than claiming there are no blockers. If
a signal was not checked, say so.
For derived signals, cite the base scope JQL and field coverage in the source
appendix instead of inventing separate support queries. Include additional JQL
only for targeted follow-up queries that were actually run.
For work item links (issuelinks), fetch linked work item status/category when
possible. If linked details are unavailable, show dependency status as unknown
rather than resolved.
Normalize
Before designing the output, create a compact renderer-independent work item
model with:
- Key, URL, summary, type, status, status category, priority
- Assignee display name or
Unassigned
- Owner status as
active, inactive, unknown, or unassigned
- Created age, updated age, resolution age when done, due date distance
- Parent/epic/workstream, sprint, estimate, components, versions, labels
- Linked work item keys, direction, link type, and linked status when available
Derived signals should stay explainable from Jira facts: done, active, not
started, stale, very stale, blocked, unowned, inactive owner, time-sensitive,
support-impacting, cross-space dependency, and missing planning data. Mark weak
text-only signals as inferred.
Dashboard Model
Create a dashboard model before rendering. Every renderer should use this same
model.
Include:
- Context metadata: title, project or space, sprint, board, filter, JQL, window,
query timestamp, and mode.
- Four top stats: committed or total scope, done or completed, active or in
progress, and needs attention.
- Scope caveat, only when sprint data is missing, mixed, stale, or blended with
recent project movement.
- Optional capacity or commitment segments, only when real data exists.
- Optional chart data, only when categories, values, units, and time ranges are
available.
- Owner load and gaps.
- Risk and attention items.
- Highest-priority work item table.
- Recently completed work.
- Source appendix with exact JQL, field coverage, assumptions, and the
composition of
Needs attention.
Do not invent data to fill the model. Empty or unsupported sections should be
omitted.
Dashboard Shape
Keep the visible dashboard simple and deterministic. When the data exists,
broadly follow this order:
Compact context header
- Show title plus project, space, board, sprint, or window metadata.
- Keep it short. Do not put queries, field coverage, or executive-summary
prose at the top.
Four-stat top bar
- Show exactly four stat values.
- Default to committed/total scope, done/completed, active/in progress, and
needs attention.
- Use work item counts when story points are unavailable.
Needs attention should combine the highest-signal risks: blocked, stale,
unassigned, time-sensitive, or unresolved linked work.
- Put secondary counts below the fold only when they change the readout.
Scope caveat, only when needed
- Use one compact caveat below the top bar when sprint data is missing,
mixed, stale, or blended with recent project movement.
- Keep it to 1 to 2 short sentences.
Capacity or commitment bar
- Render a capacity or commitment visual only when capacity, commitment, or
allocation segment data is available.
- Skip it rather than inventing capacity, segment, or buffer values.
Sprint charts
- If available, render remaining work over time as a line chart.
- Beside it, render status distribution as a pie chart or compact status
visual.
- Below those, render resolved/completed per working day as a bar chart.
- Skip any chart whose categories, values, units, or time range are missing.
Never render placeholder, sample, empty, or guessed charts.
Owner load and gaps
- Show active, stale, blocked, support-impacting, and done counts by assignee.
- Include unassigned, inactive-owner, and unknown-owner-status buckets.
- Keep it compact; prefer a small table or bar chart over per-owner cards.
Risk and attention
- Place this below owner load.
- Include only the work items most likely to need manager or lead attention.
- For each item, show key, reason, evidence, owner, age, and next question.
- Use a callout or highlighted row for the single highest delivery risk when
one stands out.
Highest-priority work item table
- Include a compact table of top sprint work items or top attention items.
- Do not render every low-signal work item by default.
Recently completed and optional detail
- If recently completed work exists, put it in a collapsed section or compact
table below the main readout.
- Workstream grouping and dependencies should appear only when they change
what the viewer should inspect next.
- Keep dependencies to unresolved or unknown-status linked work by default.
Source appendix
- Put exact JQL, query timestamps, field coverage, assumptions, and the
composition of
Needs attention at the bottom.
If the full data set is unavailable, preserve the same broad order and omit the
sections or charts that cannot be rendered honestly.
Content And Style
- Use charts and tables where they beat paragraphs.
- Follow the reference layout order when the data supports it; skip unsupported
charts instead of changing the whole page shape.
- Keep work item summaries short; avoid full descriptions unless a short excerpt
is needed to explain impact.
- Tie every recommendation or next question to work item keys or aggregate
counts.
- Separate Jira facts from derived or inferred signals.
- Use semantic tones where the renderer supports them:
success for done,
warning for stale/deadline risk, danger for blocked/overdue/severe risk,
info for caveats/linked work, and neutral for low-signal facts.
- Pair color with labels. Prefer work item key links over large buttons.
- Keep the first screen focused on status and attention, not process notes.
Renderer-Specific Style
For Cursor Canvas:
- Use Canvas components and host theme styles.
- Keep the top area compact: context header followed by exactly four stats.
- Prefer Canvas tables and charts over custom layout code.
For interactive artifacts or static HTML:
- Use a restrained dashboard layout with compact cards, tables, and simple
charts.
- Keep text readable on mobile and desktop.
- Use accessible labels for chart substitutes and status colors.
- Keep source details below the main dashboard.
For Markdown:
- Use short section headings.
- Prefer tables over paragraphs.
- Keep caveats and recommendations concise.
- Put source queries at the bottom.
Self-Check
Before returning:
- Scope came from the user or a provided URL; missing or ambiguous scope was
clarified before querying.
- The selected renderer matches the current environment's capabilities.
- Cursor Canvas instructions were used only when Cursor Canvas is available.
- If Cursor Canvas was used, the canvas imports only from
cursor/canvas.
- The top area is a compact context header followed by exactly four stats.
- There is no query list, field coverage, or executive summary above the top
bar.
- Ordinary signals were derived from the paginated scope query when possible;
targeted follow-up queries were used only when they supported a visible claim
that the scope data could not safely support.
- Counts reconcile with the queried work item set.
- Empty sections are omitted.
- Charts are rendered only when their categories, values, units, and time ranges
are available.
- Risk labels are explainable from visible Jira data.
- Source appendix includes exact JQL and field coverage for visible claims.
- No Jira write tools were used.
1---
2name: jira-sprint-dashboard
3description: >-
4 Create a visual Jira sprint dashboard from Jira project, space, sprint, board,
5 filter, JQL, work item keys, or Jira URL data. Use when the user asks for a
6 Jira sprint dashboard, standup dashboard, sprint review, delivery review,
7 engineering manager dashboard, WIP review, planning view, closeout view, or a
8 visual snapshot of Jira work that is more useful than a flat report. Use the
9 richest dashboard format supported by the current agent, such as Cursor
10 Canvas, an interactive artifact, HTML, or Markdown.
11metadata:
12 source: https://skillmd.com/skills/gabrielmoreira/jira-sprint-dashboard
13---
14
15# Jira Sprint Dashboard
16
17Build a focused dashboard that helps an engineering manager, tech lead, or
18senior engineer see current Jira work quickly enough to decide what needs
19attention. The output is a dashboard, not a prose report and not a generic
20health score.
21
22This skill is read-only by default. Do not create, update, transition, assign,
23or comment on Jira work items unless the user explicitly asks for a write action
24after reviewing the dashboard.
25
26## Output Mode
27
28Use the richest dashboard renderer supported by the current environment. The
29dashboard content, claims, counts, and source appendix must stay consistent
30across renderers; only the presentation changes.
31
32Choose the renderer in this order:
33
341. Cursor Canvas, if running in Cursor with Canvas support.
352. Interactive artifact, if the current agent supports HTML, React, or similar
36 artifact output.
373. Static HTML file, if file creation is available and useful.
384. Markdown dashboard, if no richer visual renderer is available.
395. Structured JSON plus concise summary, only if visual rendering is impossible.
40
41Do not mention that Cursor Canvas is unavailable unless the user specifically
42asked for Cursor Canvas. If the user asked for a dashboard generally, use the
43best available renderer without apologizing for the environment.
44
45## Cursor Canvas Renderer
46
47Use this section only when running in Cursor with Canvas support.
48
49Read `~/.cursor/skills-cursor/canvas/SKILL.md` before writing canvas code. If
50you need exact exports or prop shapes, read the files in
51`~/.cursor/skills-cursor/canvas/sdk/`.
52
53Canvas constraints:
54
55- Create one `.canvas.tsx` file in the Cursor canvases directory.
56- Import only from `cursor/canvas`. Do not import `react`, `CSSProperties`,
57 `JSX`, Atlaskit, or other packages.
58- Embed Jira data inline in the canvas; do not fetch from the canvas.
59- Prefer Canvas primitives such as `Stack`, `Grid`, `Card`, `Stat`, `Table`,
60 `Pill`, `Callout`, `UsageBar`, `BarChart`, `LineChart`, `PieChart`, and `Code`
61 over raw HTML.
62- Use `useHostTheme()` for custom styles. Do not hardcode hex colors, gradients,
63 box shadows, ADS variables, unsupported CSS frameworks, or `@atlaskit/*`.
64- Do not publish or share the canvas unless the user asks.
65
66## Portable Renderers
67
68Use this section when Cursor Canvas is unavailable.
69
70For an interactive artifact renderer:
71
72- Render the same dashboard model as an interactive artifact.
73- Prefer tables, compact charts, stat rows, and collapsible source details.
74- Keep interactions lightweight: filtering, expanding details, or switching
75 chart/table views is fine; do not require live Jira fetching from the artifact.
76
77For static HTML:
78
79- Create a self-contained dashboard file with embedded data.
80- Use responsive layout, accessible tables, and simple chart-like visuals when
81 chart libraries are unavailable.
82- Do not fetch Jira data from the HTML file.
83
84For Markdown:
85
86- Preserve the dashboard order.
87- Use compact tables for stats, owner load, risks, highest-priority work, and
88 source appendix.
89- Use textual chart substitutes only when they remain honest, such as counts,
90 percentages, and simple bars.
91- Avoid turning the output into a long prose report.
92
93For JSON fallback:
94
95- Return the normalized dashboard model.
96- Include a short human-readable summary with the highest-signal risks and the
97 source scope.
98
99## Get The Scope
100
101Do not guess the Jira scope. If the user does not provide a project key, space
102key, board, sprint, filter, JQL, work item keys, or Jira URL, stop and ask for
103one. A dashboard from a random visible project or guessed team context is worse
104than no dashboard.
105
106If the user gives a project or space key but no sprint, board, or filter, start
107with the Jira JQL `project` field and the user's key:
108
109```jql
110project = "SPACE_KEY" AND sprint in openSprints() ORDER BY Rank ASC
111```
112
113If the open sprint result is empty, stale, or misleading, switch to snapshot
114mode and say so in a compact caveat below the top bar:
115
116```jql
117project = "SPACE_KEY" AND statusCategory != Done ORDER BY priority DESC, updated ASC
118project = "SPACE_KEY" AND updated >= -60d ORDER BY updated DESC
119```
120
121Use a 60-day recent movement window by default unless the user asks for another
122period.
123
124Not every board supports sprints. Team-managed boards of type `simple` have no
125sprints at all: `sprint in openSprints()` returns nothing and a board-sprint
126lookup fails with "The board does not support sprints". Treat that as a signal to
127go straight to snapshot mode rather than an error worth retrying.
128
129## Query Jira
130
131Use read-only Jira search. Request only fields needed for the dashboard and
132tolerate missing fields.
133
134Useful fields: `key`, `summary`, `status`, `statusCategory`, `assignee`,
135`priority`, `issuetype`, `created`, `updated`, `resolutiondate`, `duedate`,
136`parent`, `issuelinks`, `labels`, `components`, `fixVersions`, `sprint`, and any
137available estimate/story point field.
138
139**Sprint and story points live in custom fields, which the default response omits.**
140Jira search defaults to a compact view, so a sprint dashboard built from it will
141look sprint-less even when the sprint is active. To get them, either pass
142`view: "evidence"` (or `"full"`), or request this site's `customfield_*` IDs
143explicitly — the IDs differ per Cloud site. Custom field values come back under
144`fields.customFields`, not as top-level `customfield_*` keys.
145
146Start with `maxResults: 100`. For complete sprint, board, or filter dashboards,
147paginate until the scope is complete or too large for useful work-item-level
148rendering.
149
150Default to one complete paginated scope query. Derive ordinary dashboard signals
151locally from the returned work item set instead of issuing separate JQL calls for
152each signal.
153
154Derive these locally when the scope query returned the required fields:
155
156- Recently completed from `statusCategory = Done` and `resolutiondate`.
157- Aging unfinished from `statusCategory != Done` and `updated`.
158- Unowned unfinished from `statusCategory != Done` and empty `assignee`.
159- High-priority unfinished from `statusCategory != Done` and `priority`.
160- Status or label blockers from `status`, `statusCategory`, and `labels`.
161- Owner load, stale work, due date risk, and planning gaps from the normalized
162 scope dataset.
163
164Use targeted follow-up queries only when they are needed to support a visible
165claim that cannot be derived safely from the scope data, when the scope is too
166large for useful local processing, or when the user asks for an audit-style
167dashboard with exact evidence per signal.
168
169Rule of thumb:
170
171- Small or medium sprint dashboard: prefer one full paginated scope query, with
172 at most one targeted blocker-text or dependency-status follow-up when needed.
173- Large scope dashboard: use narrower follow-up queries when pulling and
174 processing the full work-item set would be slow or low-value.
175- Evidence-heavy review: multiple focused queries are acceptable when the exact
176 JQL evidence matters more than minimizing calls.
177
178Examples of targeted follow-up queries, only when justified:
179
180- Recently completed: `<scope> AND statusCategory = Done ORDER BY resolutiondate DESC`
181- Aging unfinished: `<scope> AND statusCategory != Done AND updated <= -3d ORDER BY updated ASC`
182- Unowned unfinished: `<scope> AND statusCategory != Done AND assignee is EMPTY ORDER BY priority DESC, updated ASC`
183- High-priority unfinished: `<scope> AND statusCategory != Done AND priority in (Highest, High) ORDER BY priority DESC, updated ASC`
184- Blocked signal: `<scope> AND statusCategory != Done AND (status = Blocked OR text ~ "blocked" OR labels in (blocked, blocker)) ORDER BY priority DESC, updated ASC`
185
186Do not make negative claims such as "no blockers" or "no dependencies" unless
187the source appendix shows the query or returned field coverage that supports the
188claim. If only status and labels were checked for blockers, say that no
189status/label blockers were found rather than claiming there are no blockers. If
190a signal was not checked, say so.
191
192For derived signals, cite the base scope JQL and field coverage in the source
193appendix instead of inventing separate support queries. Include additional JQL
194only for targeted follow-up queries that were actually run.
195
196For work item links (`issuelinks`), fetch linked work item status/category when
197possible. If linked details are unavailable, show dependency status as unknown
198rather than resolved.
199
200## Normalize
201
202Before designing the output, create a compact renderer-independent work item
203model with:
204
205- Key, URL, summary, type, status, status category, priority
206- Assignee display name or `Unassigned`
207- Owner status as `active`, `inactive`, `unknown`, or `unassigned`
208- Created age, updated age, resolution age when done, due date distance
209- Parent/epic/workstream, sprint, estimate, components, versions, labels
210- Linked work item keys, direction, link type, and linked status when available
211
212Derived signals should stay explainable from Jira facts: done, active, not
213started, stale, very stale, blocked, unowned, inactive owner, time-sensitive,
214support-impacting, cross-space dependency, and missing planning data. Mark weak
215text-only signals as inferred.
216
217## Dashboard Model
218
219Create a dashboard model before rendering. Every renderer should use this same
220model.
221
222Include:
223
224- Context metadata: title, project or space, sprint, board, filter, JQL, window,
225 query timestamp, and mode.
226- Four top stats: committed or total scope, done or completed, active or in
227 progress, and needs attention.
228- Scope caveat, only when sprint data is missing, mixed, stale, or blended with
229 recent project movement.
230- Optional capacity or commitment segments, only when real data exists.
231- Optional chart data, only when categories, values, units, and time ranges are
232 available.
233- Owner load and gaps.
234- Risk and attention items.
235- Highest-priority work item table.
236- Recently completed work.
237- Source appendix with exact JQL, field coverage, assumptions, and the
238 composition of `Needs attention`.
239
240Do not invent data to fill the model. Empty or unsupported sections should be
241omitted.
242
243## Dashboard Shape
244
245Keep the visible dashboard simple and deterministic. When the data exists,
246broadly follow this order:
247
2481. **Compact context header**
249 - Show title plus project, space, board, sprint, or window metadata.
250 - Keep it short. Do not put queries, field coverage, or executive-summary
251 prose at the top.
252
2532. **Four-stat top bar**
254 - Show exactly four stat values.
255 - Default to committed/total scope, done/completed, active/in progress, and
256 needs attention.
257 - Use work item counts when story points are unavailable.
258 - `Needs attention` should combine the highest-signal risks: blocked, stale,
259 unassigned, time-sensitive, or unresolved linked work.
260 - Put secondary counts below the fold only when they change the readout.
261
2623. **Scope caveat, only when needed**
263 - Use one compact caveat below the top bar when sprint data is missing,
264 mixed, stale, or blended with recent project movement.
265 - Keep it to 1 to 2 short sentences.
266
2674. **Capacity or commitment bar**
268 - Render a capacity or commitment visual only when capacity, commitment, or
269 allocation segment data is available.
270 - Skip it rather than inventing capacity, segment, or buffer values.
271
2725. **Sprint charts**
273 - If available, render remaining work over time as a line chart.
274 - Beside it, render status distribution as a pie chart or compact status
275 visual.
276 - Below those, render resolved/completed per working day as a bar chart.
277 - Skip any chart whose categories, values, units, or time range are missing.
278 Never render placeholder, sample, empty, or guessed charts.
279
2806. **Owner load and gaps**
281 - Show active, stale, blocked, support-impacting, and done counts by assignee.
282 - Include unassigned, inactive-owner, and unknown-owner-status buckets.
283 - Keep it compact; prefer a small table or bar chart over per-owner cards.
284
2857. **Risk and attention**
286 - Place this below owner load.
287 - Include only the work items most likely to need manager or lead attention.
288 - For each item, show key, reason, evidence, owner, age, and next question.
289 - Use a callout or highlighted row for the single highest delivery risk when
290 one stands out.
291
2928. **Highest-priority work item table**
293 - Include a compact table of top sprint work items or top attention items.
294 - Do not render every low-signal work item by default.
295
2969. **Recently completed and optional detail**
297 - If recently completed work exists, put it in a collapsed section or compact
298 table below the main readout.
299 - Workstream grouping and dependencies should appear only when they change
300 what the viewer should inspect next.
301 - Keep dependencies to unresolved or unknown-status linked work by default.
302
30310. **Source appendix**
304 - Put exact JQL, query timestamps, field coverage, assumptions, and the
305 composition of `Needs attention` at the bottom.
306
307If the full data set is unavailable, preserve the same broad order and omit the
308sections or charts that cannot be rendered honestly.
309
310## Content And Style
311
312- Use charts and tables where they beat paragraphs.
313- Follow the reference layout order when the data supports it; skip unsupported
314 charts instead of changing the whole page shape.
315- Keep work item summaries short; avoid full descriptions unless a short excerpt
316 is needed to explain impact.
317- Tie every recommendation or next question to work item keys or aggregate
318 counts.
319- Separate Jira facts from derived or inferred signals.
320- Use semantic tones where the renderer supports them: `success` for done,
321 `warning` for stale/deadline risk, `danger` for blocked/overdue/severe risk,
322 `info` for caveats/linked work, and `neutral` for low-signal facts.
323- Pair color with labels. Prefer work item key links over large buttons.
324- Keep the first screen focused on status and attention, not process notes.
325
326## Renderer-Specific Style
327
328For Cursor Canvas:
329
330- Use Canvas components and host theme styles.
331- Keep the top area compact: context header followed by exactly four stats.
332- Prefer Canvas tables and charts over custom layout code.
333
334For interactive artifacts or static HTML:
335
336- Use a restrained dashboard layout with compact cards, tables, and simple
337 charts.
338- Keep text readable on mobile and desktop.
339- Use accessible labels for chart substitutes and status colors.
340- Keep source details below the main dashboard.
341
342For Markdown:
343
344- Use short section headings.
345- Prefer tables over paragraphs.
346- Keep caveats and recommendations concise.
347- Put source queries at the bottom.
348
349## Self-Check
350
351Before returning:
352
353- Scope came from the user or a provided URL; missing or ambiguous scope was
354 clarified before querying.
355- The selected renderer matches the current environment's capabilities.
356- Cursor Canvas instructions were used only when Cursor Canvas is available.
357- If Cursor Canvas was used, the canvas imports only from `cursor/canvas`.
358- The top area is a compact context header followed by exactly four stats.
359- There is no query list, field coverage, or executive summary above the top
360 bar.
361- Ordinary signals were derived from the paginated scope query when possible;
362 targeted follow-up queries were used only when they supported a visible claim
363 that the scope data could not safely support.
364- Counts reconcile with the queried work item set.
365- Empty sections are omitted.
366- Charts are rendered only when their categories, values, units, and time ranges
367 are available.
368- Risk labels are explainable from visible Jira data.
369- Source appendix includes exact JQL and field coverage for visible claims.
370- No Jira write tools were used.