| name | jira-setup |
| description | Ensure the mcp-atlassian MCP server is registered and resolve Jira defaults (URL, auth, project, board) so the jira-ticket skill has everything it needs. Idempotent — re-runs only prompt for whatever is missing. Use when the user wants to set up or reconfigure Jira, install / register mcp-atlassian, pick a default project or board, or persist Jira env vars to their shell rc. Also invoked automatically by the jira-ticket skill when required env vars are missing. Triggers on: "/jira-setup", "set up jira", "configure jira", "install mcp-atlassian", "pick jira project", "pick jira board".
|
Jira Setup
Walk through the steps below and prompt only for what's missing. Re-running this is safe — values that are already set are reported and skipped.
Step 0 — Ensure the mcp-atlassian MCP server is registered
This skill (and jira-ticket) drive Jira through the mcp__mcp-atlassian__* tools. Check whether mcp__mcp-atlassian__jira_create_issue is visible in your tool list.
-
If it's visible, the server is registered. Skip to Step 1.
-
If it's missing, hand off to the setup-mcp skill, which is the bounded installer for this plugin's MCP servers — mcp-atlassian is already in its manifest. Invoke setup-mcp via the Skill tool with mcp-atlassian (e.g. /setup-mcp mcp-atlassian). It shows the proposed config, asks for scope, registers the server, and verifies the uvx prerequisite. Because MCP servers only load at startup, it will tell the user to restart Claude Code — so once it returns, tell the user to restart and re-run /jira-setup (or /jira-ticket), then stop. The tools won't be available until the restart.
If the setup-mcp skill isn't available in this session for any reason, fall back to instructing the user: "Run /setup-mcp mcp-atlassian, restart Claude Code, then re-run /jira-setup."
Step 1 — Validate credentials
Run this exact command:
echo "JIRA_URL=${JIRA_URL:-NOT_SET}" && \
echo "JIRA_USERNAME=${JIRA_USERNAME:-NOT_SET}" && \
echo "JIRA_API_TOKEN=${JIRA_API_TOKEN:+SET}" && \
echo "JIRA_PERSONAL_TOKEN=${JIRA_PERSONAL_TOKEN:+SET}"
JIRA_URL is always required. For auth, either the Cloud combo (JIRA_USERNAME + JIRA_API_TOKEN) or the Server / Data Center PAT (JIRA_PERSONAL_TOKEN) must be set. If neither is complete, instruct the user and stop. Do not proceed.
JIRA_URL is NOT_SET → "Set your Jira workspace URL, e.g. export JIRA_URL=https://yourworkspace.atlassian.net"- Cloud (Atlassian-hosted,
*.atlassian.net): set both
- Server / Data Center (self-hosted): set just
JIRA_PERSONAL_TOKEN → create a Personal Access Token in your Jira profile, then export JIRA_PERSONAL_TOKEN=<your-pat>. JIRA_USERNAME is optional here, but if you set it the jira-ticket skill will use it for the assignee / reporter on created tickets.
The mcp-atlassian server reads these from the shell that launched Claude Code, so encourage the user to add the relevant exports to their ~/.zshrc or ~/.bashrc and restart Claude Code so the server picks them up.
Step 2 — Resolve project (auto-discover if needed)
Check $JIRA_PROJECT_KEY:
echo "JIRA_PROJECT_KEY=${JIRA_PROJECT_KEY:-NOT_SET}"
If set, report the resolved value and move on. If NOT_SET:
- Call
mcp__mcp-atlassian__jira_get_all_projects to list available projects.
- Present them to the user as a numbered list of
KEY — Name (e.g., 1. PROJ — My Project). If there are many, group/sort sensibly.
- Ask the user to pick one.
- Treat the picked key as the resolved
$JIRA_PROJECT_KEY for the rest of this run.
- Offer to persist: "Save
JIRA_PROJECT_KEY=<KEY> to your shell rc so you don't have to pick again next time? (yes/no)"
- If yes → follow the Persisting to shell rc procedure below with
VAR_NAME=JIRA_PROJECT_KEY and VAR_VALUE=<picked key>.
Step 3 — Resolve board (optional, for sprint placement)
This is only needed if the user wants to place tickets into sprints. If this skill is being invoked from jira-ticket and no sprint argument was passed, you can skip this step. Otherwise, offer to set it up so future sprint placements work without prompting.
Check $JIRA_BOARD_ID:
echo "JIRA_BOARD_ID=${JIRA_BOARD_ID:-NOT_SET}"
If set, report it and move on. If NOT_SET:
- Call
mcp__mcp-atlassian__jira_get_agile_boards filtered by the resolved project key.
- Present them to the user as a numbered list of
ID — Name (Type) (e.g., 1. 6 — Engineering Scrum (scrum)).
- Ask the user to pick one.
- Treat the picked id as the resolved
$JIRA_BOARD_ID for the rest of this run.
- Offer to persist: "Save
JIRA_BOARD_ID=<ID> to your shell rc? (yes/no)"
- If yes → follow the Persisting to shell rc procedure below with
VAR_NAME=JIRA_BOARD_ID and VAR_VALUE=<picked id>.
If the project has no boards, tell the user and skip — sprint placement won't be available until a board exists.
Persisting to shell rc
When the user agrees to persist a value, detect the right rc file, check whether the export is already present, and append only if missing. Run:
VAR_NAME=JIRA_PROJECT_KEY
VAR_VALUE=PROJ
case "$(basename "${SHELL:-}")" in
zsh) RC=~/.zshrc ;;
bash) RC=~/.bashrc ;;
*) RC=~/.profile ;;
esac
touch "$RC"
if grep -qE "^export[[:space:]]+${VAR_NAME}=" "$RC"; then
echo "${VAR_NAME} is already exported in $RC — leaving it alone. Update it manually if you want a different value."
else
echo "export ${VAR_NAME}=${VAR_VALUE}" >> "$RC"
echo "Added 'export ${VAR_NAME}=${VAR_VALUE}' to $RC"
fi
Then tell the user: "The current session is already using <value>. New shells will pick it up automatically; for an existing shell, run source $RC."
Final summary
After all applicable steps complete, print a one-line readiness summary so the user knows what's wired up:
Jira ready ✓
URL: <JIRA_URL>
Auth: <Cloud (JIRA_USERNAME) | Server/DC (JIRA_PERSONAL_TOKEN)>
Project: <JIRA_PROJECT_KEY>
Board: <JIRA_BOARD_ID or "not set — sprint placement unavailable">
If anything required is still missing, list what's missing and how to set it instead of claiming readiness.
