Jira Pm

Mode: Cognitive/Prompt-Driven - No standalone utility script; use via agent context.

Required:

• JIRA_URL - Base URL of your Jira instance (e.g., https://yourcompany.atlassian.net)
• JIRA_API_TOKEN - API token for authentication (generate at Jira Account Settings -> Security -> API Tokens)
• JIRA_USER_EMAIL - Email address associated with the API token

Optional:

• JIRA_DEFAULT_PROJECT - Default project key for operations (e.g., PROJ)
• JIRA_API_VERSION - API version (default: 3)

Issues

Projects

Sprints

Comments

Common Workflows

Issue Creation: List projects -> Get project info -> Create issue -> Add comment

Sprint Management: Get active sprint -> List sprint issues -> Update issue status -> Add comments

Issue Search: Use JQL for targeted searches -> Retrieve issue details -> Update issues

Primary Agent: planner

• Use Case: Project backlog management, sprint planning, requirement tracking
• Common Operations: Create issues, update priorities, manage sprints

Secondary Agent: developer

• Use Case: Issue tracking during development, status updates
• Common Operations: Search assigned issues, transition issues, add comments

Common Error Scenarios

• Authentication Errors: Missing or invalid API token, expired credentials, insufficient permissions
• Rate Limiting: Too many API requests in short period, implement exponential backoff
• Invalid Inputs: Non-existent project keys, invalid issue types, invalid transition IDs
• Network Errors: Connection timeouts, unreachable Jira instance

1. Use JQL Efficiently: Craft precise JQL queries to reduce result sets and API calls
2. Cache Metadata: Store project keys, issue types, and transitions locally
3. Verify Before Create: Always verify project and issue type before creating issues
4. Use Transitions: Respect workflow states when changing issue status
5. Batch Operations: Group related API calls when possible
6. Handle Errors Gracefully: Provide clear error messages and recovery suggestions
7. Respect Rate Limits: Implement backoff strategies for high-volume operations

Context Optimization

1. Lazy Loading: Only load issue details when explicitly requested
2. Field Selection: Request only necessary fields from Jira API
3. Caching: Store frequently accessed metadata (projects, issue types)
4. Streaming: Not supported - all responses are complete payloads
5. Pagination: Automatically handle large result sets

Context Savings: 90%+ compared to loading full Jira MCP server

Jira REST API Endpoints Used

• /rest/api/3/search - JQL search
• /rest/api/3/issue/{issueKey} - Get/update issue
• /rest/api/3/issue - Create issue
• /rest/api/3/issue/{issueKey}/transitions - Transition issue
• /rest/api/3/project - List projects
• /rest/api/3/project/{projectKey} - Get project details
• /rest/agile/1.0/board/{boardId}/sprint - Get sprints
• /rest/agile/1.0/sprint/{sprintId}/issue - Get sprint issues
• /rest/api/3/issue/{issueKey}/comment - Get/add comments

See https://developer.atlassian.com/cloud/jira/platform/rest/v3/ for full reference.

Iron Laws

1. ALWAYS verify the project key exists before creating an issue — Jira silently ignores invalid project keys in some API versions, creating orphaned issues or returning cryptic errors.
2. NEVER use PUT /issue to change status — Jira status changes must go through valid workflow transitions via POST /issue/{key}/transitions; direct field updates bypass workflow validators and automation rules.
3. ALWAYS use JQL for issue searches rather than fetching all issues and filtering client-side — returning all issues wastes quota and causes timeouts on projects with thousands of tickets.
4. NEVER create duplicate issues without first searching for existing ones — duplicate tickets fragment tracking, confuse assignees, and produce misleading velocity metrics.
5. ALWAYS include summary, issuetype, and project fields when creating an issue — these three fields are the minimum required by Jira Cloud REST API v3; missing any produces a 400 error.

Anti-Patterns

Memory Protocol (MANDATORY)

Before starting: Read .claude/context/memory/learnings.md

After completing:

• New pattern -> .claude/context/memory/learnings.md
• Issue found -> .claude/context/memory/issues.md
• Decision made -> .claude/context/memory/decisions.md

ASSUME INTERRUPTION: If it is not in memory, it did not happen.