Jira Task Context

Use this skill when the user gives a Jira URL/key or asks to start from a Jira task. The goal is to collect the root task details and linked Jira issue context before doing the rest of the work.

What to do

1. Extract the Jira issue key and base URL from the user-provided URL/key.
2. Fetch the root issue details.
3. Follow linked Jira issues at least one hop by default:
   • explicit Jira issue links (blocks, is blocked by, relates to, etc.)
   • parent issue
   • subtasks
   • Jira issue keys/URLs referenced in the description or recent comments
4. Extract important non-Jira links from the root and followed Jira issues, especially:
   • Confluence / Atlassian wiki pages
   • Figma design, FigJam, prototype, or file links
   • product specs, API docs, screenshots, videos, or other implementation references
5. Follow those external links read-only when access/tools are available:
   • For Confluence pages, use web_fetch when authenticated access works; summarize requirements, decisions, constraints, screenshots/assets mentioned, and open questions.
   • For Figma links, load the figma skill and use the project-enabled native Figma MCP to inspect the referenced file/frame/prototype. Discover tools through codemode (or tool_search when available) and load the server's required guidance before get_design_context; do not use the removed personal-token extension tools or the Xcode-specific project_tool_search. Keep this context lookup read-only even when server guidance describes implementation or Figma writes. If MCP access is unavailable, try web_fetch for page metadata and clearly report that visual/design details could not be inspected.
   • Do not mutate Confluence, Figma, Jira, or any external system.
   • Keep traversal bounded to links directly referenced by the collected Jira issues unless the user asks for deeper discovery.
6. Summarize the collected Jira and external-artifact context for the user before moving on.

Preferred command

From any working directory, run the bundled helper script:

python3 ~/.pi/agent/skills/jira-task-context/scripts/jira_issue_context.py "<jira-url-or-key>" --depth 1 --comments 5

Useful variants:

# Root issue only
python3 ~/.pi/agent/skills/jira-task-context/scripts/jira_issue_context.py "<jira-url-or-key>" --depth 0

# Follow linked issues two hops, with a strict cap
python3 ~/.pi/agent/skills/jira-task-context/scripts/jira_issue_context.py "<jira-url-or-key>" --depth 2 --max-issues 30

# If the user gave only a key, provide or rely on JIRA_BASE_URL
JIRA_BASE_URL="https://your-domain.atlassian.net" python3 ~/.pi/agent/skills/jira-task-context/scripts/jira_issue_context.py "ABC-123"

# Machine-readable output
python3 ~/.pi/agent/skills/jira-task-context/scripts/jira_issue_context.py "<jira-url-or-key>" --format json

Authentication

Do not ask the user to paste tokens into chat and never print token values.

Supported environment variables:

# Jira Cloud basic auth
export JIRA_EMAIL="name@example.com"
export JIRA_API_TOKEN="..."

# Jira Data Center / PAT bearer auth
export JIRA_BEARER_TOKEN="..."
# or
export JIRA_PAT="..."

# Required only when the user supplies an issue key instead of a full URL
export JIRA_BASE_URL="https://your-domain.atlassian.net"

If authentication fails, tell the user which variable names are needed, not the secret values.

Output expectations

After fetching, present a concise context summary containing:

• root issue key, title, URL, type, status, priority, assignee, reporter
• description / acceptance criteria / important comments
• linked issues with relation, status, and why they matter
• blockers/dependencies inferred from issue link relations
• Confluence/spec links followed, with relevant requirements or decisions found
• Figma/design links followed, with referenced file/frame/prototype and visual requirements found
• inaccessible or unfollowed links, with the reason and whether they block implementation
• open questions or missing information

Then continue with the user's requested flow.

Safety and privacy

• Jira data may contain confidential product/customer information. Only fetch and summarize what is needed for the task.
• Keep max issue limits reasonable (--max-issues) to avoid accidentally crawling too much Jira content.
• Do not mutate Jira issues unless the user explicitly asks. This skill is read-only by default.