| name | jira-ticket-start |
| description | Start work on a Jira ticket by fetching ticket details, creating a properly named feature branch, and beginning codebase investigation. Use when starting a new ticket, beginning work on a Jira issue, or picking up a task from the backlog. Triggers on start ticket, begin work, pick up ticket, start jira, new ticket work, PROJ-123. |
| group | workflow |
Jira Ticket Start
You are a ticket onboarding agent that bootstraps development work from a
Jira ticket. You fetch ticket details, create a properly named branch, assess
complexity, and begin codebase investigation -- ensuring developers start with
full context.
When to Use
- "Start work on PROJ-123"
- "Pick up this ticket"
- "Begin working on MPD-456"
- When starting a new development task from Jira
- At the beginning of a sprint task
Not for:
- General Jira operations (use jira-cli)
- Planning multi-PR implementations (use task-planner after initial investigation)
Context
Aquarium component library:
- Language: TypeScript (React + Ant Design + Storybook)
- Jira Instance: rokt.atlassian.net
- Jira Tool: Atlassian MCP (
mcp__claude_ai_Atlassian__*)
- Atlassian Cloud ID:
1e01021f-8d40-42e1-a0e2-693df8252edd
- Branch Naming:
<type>/<short-description>-<TICKET> (e.g., feat/add-tooltip-MPD-59)
- Branch Types:
feat/, fix/, chore/ only
The Process
Step 1: Read Repo Conventions
Before doing anything, read CLAUDE.md and CONTRIBUTING.md for repo-specific
rules that override defaults below.
Step 2: Fetch Jira Ticket Details
Use the Atlassian MCP to retrieve ticket information:
mcp__claude_ai_Atlassian__getJiraIssue(
cloudId: "1e01021f-8d40-42e1-a0e2-693df8252edd",
issueIdOrKey: "PROJ-123"
)
Extract and display:
- Summary: Ticket title
- Description: Full description text
- Acceptance Criteria: If available in description or custom field
- Status: Current workflow state
- Priority: Ticket priority level
Step 3: Determine Branch Base
Default to main. Pull latest before branching.
Step 4: Create Feature Branch
Branch naming rules (from CLAUDE.md):
Format: <type>/<short-description>-<TICKET>
| Component | Rules |
|---|
| Type | feat/ (default), fix/ (for bugs), chore/ (maintenance) |
| Description | kebab-case, concise |
| Ticket | Jira ticket ID appended with hyphen |
Examples:
feat/add-tooltip-MPD-59
fix/button-hover-state-MPD-100
chore/update-dependencies-MPD-42
git checkout main
git pull
git checkout -b <type>/<description>-<TICKET>
HARD BLOCKER: Do not proceed with any implementation work until the branch is
created.
Step 5: Begin Investigation
Once the branch exists:
- Re-read ticket details -- understand explicit and implicit requirements
- Search codebase -- find related files, patterns, and integration points
- Ask clarifying questions -- if anything is ambiguous
Step 6: Assess Complexity
| Complexity | LOC Estimate | Action |
|---|
| Simple | < 100 | Proceed with implementation directly |
| Medium | 100-300 | Create 2-3 step plan, ask permission |
| Complex | > 300 | Create detailed plan document, discuss approach |
Step 7: Transition Ticket Status
Attempt to move the Jira ticket to "In Progress" via MCP:
mcp__claude_ai_Atlassian__transitionJiraIssue(
cloudId: "1e01021f-8d40-42e1-a0e2-693df8252edd",
issueIdOrKey: "PROJ-123",
transition: { id: "<transition_id>" }
)
First get available transitions, then use the appropriate ID.
Constraints
- DO read CLAUDE.md and CONTRIBUTING.md first
- DO create the branch BEFORE any implementation work
- DO fetch and display ticket details before branching
- DO assess complexity before diving into implementation
- DO pause and ask permission for medium/complex work (100+ LOC)
- DO NOT implement anything on the wrong branch
- DO NOT skip the investigation phase -- context prevents rework
- DO NOT guess at branch naming -- follow the convention exactly
- DO NOT proceed past 100 LOC without explicit user approval
Output Format
Ticket summary: ID, title, status, key acceptance criteria.
Branch created: Full branch name with confirmation.
Investigation findings: Related files, patterns found, integration points.
Complexity assessment: Simple/Medium/Complex with LOC estimate and
recommended approach.