Jira Issues

Jira: Issue Search

Read-only. Run from this skill's directory:

python3 ../jira/scripts/jira_tool.py search --jql "assignee = currentUser() AND updated <= -14d" \
  [--max_results 100] [--only summary,status,priority] [--fields customfield_10056]

(First-time setup, once per environment: pip install -r ../jira/requirements.txt.)

--jql is required. Prints one JSON document: structured issue results matching the query. key, url, custom_fields, and blocked are always present; every other field is controlled by --only. Never invent or fabricate issue data -- everything you state must come from this JSON.

• --only (optional, comma-separated named fields, e.g. summary,status,priority,labels,description) asks for exactly the fields you need instead of everything -- the main lever for keeping bulk results token-cheap. Omit it for the default set: everything except description and the time-tracking fields (description is free text that can run long, and is rarely needed to scan many issues at once -- pass it explicitly, e.g. --only summary,description, when you actually need it). Valid names: summary, status, priority, issue_type, assignee, reporter, updated, created, due_date, labels, links, description, components, subtasks, original_estimate_seconds, time_spent_seconds, remaining_estimate_seconds.
• --fields (optional, comma-separated) requests extra raw Jira field IDs in addition to --only -- use it for instance-specific custom fields, e.g. a "Figma Link" field, always surfaced in custom_fields. Discover its ID first by running list_fields (via the jira skill's CLI, python3 ../jira/scripts/jira_tool.py list_fields); never guess a customfield_NNNNN id.
• --only's field names and --jql's field names are different vocabularies -- don't mix them up. --only uses this skill's own snake_case names (due_date, issue_type); --jql uses Jira's own JQL field names, which differ: due (not due_date), issuetype (not issue_type), reporter for "who filed this" (not creator -- don't invent an alternate field name just because a query returns nothing). Never write a status literal into --jql (e.g. status = 'Pending') unless you've actually seen that exact status name in a prior result -- a plausible guess doesn't error, it just silently returns zero misleading results. If a query built with the right, known field names still returns nothing, say exactly what you searched and ask, rather than silently swapping in a different field name and retrying.
• For subtask/description/component questions ("which tasks have no subtasks", "does this need frontend or backend work"), request the relevant fields via --only and reason over them yourself -- there is no separate classification tool, because "frontend" vs. "backend" isn't a fixed Jira field; it's inferred from this data.

If the result contains "error", tell the user what went wrong in plain language (invalid JQL, permission denied, etc.) instead of retrying silently or fabricating a result.

See ../jira/README.md for architecture details and the full environment-variable table.