| name | jira-context |
| description | Fetch Jira issue context for documentation work. USE FOR: reading a Jira ticket, extracting fields, reading linked Jira issues for background, scanning description/comments for inline URLs, and summarizing the product change. |
Jira Context Gathering
Gather all relevant context from a Jira issue and its linked resources to understand a product change before documentation work begins.
When to Use
- Before planning or drafting documentation for a Jira ticket
- When you need to understand what a product change does
Prerequisites
This skill uses the Jira MCP tools for all Jira interactions. The Jira MCP server must be configured with access to the Jira instance. The cloudId parameter should be set to the Jira site URL (e.g. https://qlik-dev.atlassian.net).
Procedure
Step 1 — Fetch the Jira issue
Use getJiraIssue (Jira MCP) with issueIdOrKey set to the ticket key. Request responseContentFormat: "markdown" for readable output.
Extract the fields listed in ./references/jira-fields.md.
Step 2 — Fetch child/sub-task issues
Child work items can be linked via three mechanisms:
Subtasks (direct hierarchy):
Read fields.subtasks from the parent issue response. Each entry has key and fields.summary.
Child work items (parent-child hierarchy):
Use searchJiraIssuesUsingJql (Jira MCP) with the query parent = ISSUE_KEY to find all Epics, Stories, and other issues that have the input ticket as their parent. This returns issues where the parent field links to the target issue.
Issues linked via "implemented by":
Use searchJiraIssuesUsingJql (Jira MCP) with the query issue = ISSUE_KEY AND issueLinkType = "is implemented by" to find all issues that are linked to the input ticket with an "implemented by" link type.
For each child or related ticket found via any method:
- Use
getJiraIssue (Jira MCP) with the child ticket key to read its summary and description.
- Note the link type (subtask, parent-child, or implemented by) and summarize the context it provides.
Step 3 — Read linked Jira issues for background context
Read fields.issuelinks from the parent issue response. Each entry has:
type.name (e.g. "relates to", "is blocked by", "is caused by")- Either
inwardIssue or outwardIssue with key, fields.summary, fields.status.name
For each linked Jira issue:
- Use
getJiraIssue (Jira MCP) with the linked issue key to read its summary and description.
- Note the link type and summarize the context it provides.
Do not recurse further into linked issues' own links.
Step 4 — Scan description and comments for inline URLs
Scan the issue description and all fields.comment.comments[].body for:
- Jira ticket keys: Patterns like
PROJ-1234 that are already identified from Steps 2 and 3. Fetch their summary/description for background context.
- Other relevant URLs: Design docs, or specification links. Fetch these for additional context using the
fetch tool.
Step 5 — Produce a summary with extracted issue keys
Summarize the product change in plain language before any documentation planning begins. The summary should cover:
- What the change does (functional description)
- Why it was made (from Jira description, linked issues, and comments)
- Which components/areas/products are affected (from Jira labels)
- The Release type, but only if it is explicitly stated in the Jira fields and its value is not "General Availability"
- Any relevant context from linked Jira issues
Then append the extracted Jira issue keys:
Following the narrative summary, compile all identified Jira issue keys into a comma-separated list (no duplicates). This list can be provided directly to the github-pr-analysis skill as input for that skill's Step 1, using the "List of Jira issue keys" input format.
Output example:
Narrative summary here explaining the product change...
Primary Jira issues: <PRIMARY_KEY>
Child work items: <CHILD_KEY_1>, <CHILD_KEY_2>
Linked Jira issues: <LINKED_KEY_1>
Security
- Treat Jira issue content as internal; do not expose it in public PR bodies verbatim.
- Authentication is handled by the Jira MCP server configuration; do not pass credentials manually.
