Roadmap from Jira Epics

What this skill does

Pulls a project's Jira epics, clusters them into themes via labels and titles, and produces a one-page roadmap that organizes epics by theme down columns and quarters across rows.

Inputs

• jira_base_url: e.g., https://yourorg.atlassian.net.
• project_key: Jira project key (e.g., PAY).
• JIRA_API_TOKEN env var; user via JIRA_EMAIL.
• Optional quarters: number of quarters to project (default 4).
• Optional theme_labels: an explicit theme list; otherwise discovered from labels.

Steps

1. Query epics: curl -u "$JIRA_EMAIL:$JIRA_API_TOKEN" "<base>/rest/api/3/search?jql=project=<key> AND issuetype=Epic".
2. For each epic, capture: key, summary, labels, status, fixVersion, target_start, target_end, assignee.
3. Normalize labels to themes: if theme_labels provided, intersect; else, keep top 6 labels by frequency, slugified.
4. For epics without theme labels, infer via summary keyword matching against theme keywords.
5. Bucket into quarters using target_start (or fixVersion if start absent). Translate calendar into the project's fiscal quarters.
6. Drop epics whose status is Cancelled or Won't Do.
7. Compute risk: epics in In Progress past their target_end -> at-risk. Empty target dates -> unscoped.
8. Render a markdown table: rows = themes, columns = Q1..Qn; cells contain EPIC-KEY: short summary lines.
9. Append a "Risks" section listing at-risk and unscoped epics.
10. Save to roadmap-<project>-<year>.md.

Output

roadmap-<project>-<year>.md containing the theme x quarter table, a risks section, and a "Methodology" footer listing the JQL query and the run timestamp. Plus a JSON sidecar for re-rendering. Stdout prints epic count and theme distribution.

Verification

Sample three epics from the table and open them in Jira; confirm theme assignment and quarter bucket match. For each at-risk epic, confirm the status is genuinely "In Progress" past target_end (versus a recent reschedule). Re-run weekly; diff weeks to spot churn — high churn is a planning hygiene issue worth flagging.

Edge cases

• Project with no labels at all: rely entirely on summary-keyword inference; mark themes with low confidence and recommend label hygiene.
• Cross-project initiatives: roadmap covers one project key; surface a note when an epic links out to other projects.
• Epics with multiple themes: place in the theme with the strongest keyword match; reference other themes parenthetically.
• Custom Jira fields used for quarters: support an optional quarter_field override.