Agent Skills: fetch-jira-ticket

Fetch a Jira ticket using the Atlassian MCP plugin and extract key details into a structured summary.

Rules

• Input Validation: A Jira ticket ID must be provided (e.g., JED-1204, PROJ-123). If not provided, STOP and ask the user for the ticket ID.

• Atlassian Cloud ID: Use cloud ID 8e5e4495-e477-4a3f-af21-f2ff97eae820 (dekeo.atlassian.net) for all requests. If this fails, call mcp__plugin_atlassian_atlassian__getAccessibleAtlassianResources to discover the correct cloud ID.

• Fetch via Atlassian MCP Plugin: Use mcp__plugin_atlassian_atlassian__getJiraIssue with the following parameters:

  • cloudId: 8e5e4495-e477-4a3f-af21-f2ff97eae820
  • issueIdOrKey: the ticket ID
  • responseContentFormat: markdown
  • expand: renderedFields,names
  • fields: ["summary", "description", "status", "priority", "assignee", "reporter", "created", "updated", "resolutiondate", "resolution", "labels", "sprint", "parent", "duedate", "comment", "attachment", "customfield_10020", "customfield_10015", "customfield_10037", "customfield_10036"]

• Handling Large Responses: The API response may be too large to display inline. If the result is saved to a file, use jq via Bash to extract fields. Run two jq commands:

  1. Metadata extraction:

     cat <result_file> | jq '{
       summary: .issues.nodes[0].fields.summary,
       status: .issues.nodes[0].fields.status.name,
       resolution: .issues.nodes[0].fields.resolution.name,
       priority: .issues.nodes[0].fields.priority.name,
       assignee: .issues.nodes[0].fields.assignee.displayName,
       reporter: .issues.nodes[0].fields.reporter.displayName,
       created: .issues.nodes[0].fields.created,
       updated: .issues.nodes[0].fields.updated,
       resolutiondate: .issues.nodes[0].fields.resolutiondate,
       labels: .issues.nodes[0].fields.labels,
       duedate: .issues.nodes[0].fields.duedate,
       parent_key: .issues.nodes[0].fields.parent.key,
       parent_summary: .issues.nodes[0].fields.parent.fields.summary,
       parent_status: .issues.nodes[0].fields.parent.fields.status.name,
       size: .issues.nodes[0].fields.customfield_10037,
       team: .issues.nodes[0].fields.customfield_10036,
       attachment_names: [.issues.nodes[0].fields.attachment[]?.filename],
       comment_count: .issues.nodes[0].fields.comment.total
     }'
     

  2. Description extraction:

     cat <result_file> | jq '.issues.nodes[0].fields.description'
     

• Comment Extraction: Comments may be in ADF (Atlassian Document Format) instead of plain text. To extract readable text from ADF comments, use this jq command on the comments file:

  cat <comments_file> | jq -r '
  def extract_text:
    if type == "object" then
      if .type == "text" then .text // ""
      elif .type == "hardBreak" then "\n"
      elif .type == "heading" then
        "\n" + (if .attrs.level == 1 then "# " elif .attrs.level == 2 then "## " elif .attrs.level == 3 then "### " else "" end) + ([.content[]? | extract_text] | join("")) + "\n"
      elif .type == "paragraph" then ([.content[]? | extract_text] | join("")) + "\n"
      elif .type == "bulletList" then ([.content[]? | extract_text] | join(""))
      elif .type == "orderedList" then ([.content[]? | extract_text] | join(""))
      elif .type == "listItem" then "- " + ([.content[]? | extract_text] | join(""))
      elif .type == "codeBlock" then "```\n" + ([.content[]? | extract_text] | join("")) + "\n```\n"
      elif .type == "mediaInline" or .type == "mediaSingle" or .type == "media" then "[Image]"
      elif .type == "inlineCard" then "[Link: " + (.attrs.url // "unknown") + "]"
      else ([.content[]? | extract_text] | join(""))
      end
    elif type == "array" then [.[] | extract_text] | join("")
    else "" end;
  .[] | "### " + .author + " - " + .created + "\n\n" + (.body | extract_text) + "\n---"
  '
  

  First try to get comments from the main result file:

  cat <result_file> | jq '[.issues.nodes[0].fields.comment.comments[] | {author: .author.displayName, created: .created, body: .body}]'
  

  If the output is too large, it will be saved to a separate file — run the ADF extraction on that file.

• Extract Key Details: Parse and present the following fields from the ticket:

  1. Title: The ticket summary/title.
  2. Description: Full description text. Preserve formatting where possible.
  3. Acceptance Criteria: Extract if present (often in description or a custom field).
  4. Implementation Plan: Extract if present (often in description or a custom field).
  5. Metadata: Include:
     • Status
     • Resolution
     • Priority
     • Assignee
     • Reporter
     • Created date
     • Updated date
     • Resolution Date
     • Labels (if any)
     • Sprint (if applicable)
     • Parent Epic (key + summary + status)
     • Due Date
     • Size (if applicable)
     • Team (if applicable)
  6. Attachments: List all attachment filenames.
  7. Comments: List all comments with author and timestamp, with full text content extracted from ADF.

• Attachments and Images: Do NOT fetch attachments or images from Jira. Instead:

  • For attachments, first check if you can find them in the project, if not, add placeholder links (e.g., [Attachment: <filename>])
  • For images embedded in comments, use [Image] placeholder
  • Do not download or embed attachment or image content

• Output Format: Present the ticket information in a clean markdown structure saved to TICKET_INFO_<Ticket_num>.md in the project root (e.g., JED-1204 as TICKET_INFO_1204.md):

  # <Ticket ID>: <Title>
  
  ## Metadata
  - **Status**: ...
  - **Resolution**: ...
  - **Priority**: ...
  - **Assignee**: ...
  - **Reporter**: ...
  - **Created**: ...
  - **Updated**: ...
  - **Resolution Date**: ...
  - **Labels**: ...
  - **Sprint**: ...
  - **Parent Epic**: <KEY> - <Summary> (<Status>)
  - **Due Date**: ...
  - **Size**: ...
  - **Team**: ...
  
  ## Description
  <description content>
  
  ## Acceptance Criteria
  <acceptance criteria if found, otherwise "Not specified">
  
  ## Implementation Plan
  <implementation plan if found, otherwise "Not specified">
  
  ## Attachments
  - [Attachment: <filename>]
  
  ## Comments
  ### <Author> - <Timestamp>
  <comment content>
  

• Missing Fields: If a field is empty or not present, indicate "Not specified" rather than omitting the section.

• No Hallucinations: Only include information that exists in the ticket. Do not infer or guess content.

• If the MCP call fails: Report the error clearly. Common issues:

  • Wrong cloud ID → call getAccessibleAtlassianResources to discover correct one
  • Permission denied → inform the user to check their Atlassian plugin authorization