JIRA Assistant

This hub routes requests to specialized JIRA skills. It does not execute JIRA operations directly—it helps find the right skill.

Risk Levels

Risk Legend: - Safe, read-only | ! Caution, modifiable | !! Warning, destructive but recoverable | !!! Danger, irreversible

Quick Reference

Risk Legend: - Read-only/safe | ⚠️ Has destructive ops (confirm) | ⚠️⚠️ High-risk (confirm + dry-run)

Routing Rules

1. Explicit skill mention wins - If user says "use jira-agile", use it
2. Entity signals - Issue key present → likely jira-issue or jira-lifecycle
3. Quantity determines bulk - 50+ issues → jira-bulk (consider bulk for 10+, required for 50+)
4. Keywords drive routing:
   • "show", "view", "display", "get", "retrieve", "see", "details", "look up", "check" (with issue reference) → jira-issue
   • "create", "update", "delete" (single issue) → jira-issue
   • "search", "find", "JQL", "filter" → jira-search
   • "sprint", "epic", "backlog", "story points", "subtask" → jira-agile
   • "transition", "move to", "assign", "close", "version", "release", "component", "resolve", "reopen", "archive" → jira-lifecycle
   • "comment", "attach", "watch", "notify", "notification", "activity", "history", "changelog" → jira-collaborate
   • "link", "blocks", "depends on", "clone", "dependency graph", "blocker chain" → jira-relationships
   • "log time", "worklog", "estimate", "time report", "timesheet" → jira-time
   • "service desk", "SLA", "customer", "request", "queue", "approval", "knowledge base", "asset" → jira-jsm
   • "branch name", "commit", "PR" → jira-dev
   • "custom field", "field ID" → jira-fields
   • "cache", "warm cache", "project discovery", "diagnostics", "performance", "request batching" → jira-ops
   • "permissions", "project settings", "automation", "automation rule", "users", "groups", "notifications", "screens", "issue types", "workflows", "notification scheme", "permission scheme" → jira-admin

Negative Triggers

When to Clarify First

Ask the user before routing when:

• Request matches 2+ skills with similar likelihood
• Request is vague or could be interpreted multiple ways
• Destructive operations are implied

Disambiguation Table

Disambiguation Examples

"Show me the sprint" Could mean:

1. Sprint metadata (dates, goals, capacity) → jira-agile
2. Issues in the current sprint → jira-search

Ask: "Do you want sprint details or the issues in the sprint?"

"Update the issue" Could mean:

1. Change fields on one issue → jira-issue
2. Transition status → jira-lifecycle
3. Update multiple issues → jira-bulk

Ask: "What would you like to update - fields, status, or multiple issues?"

"Create an issue in the epic" Context determines:

• Epic context explicit → jira-agile
• Just issue creation → jira-issue

"Find all P1 bugs and close them" Multi-step workflow:

1. First search with jira-search to find issues
2. Then confirm count before using jira-bulk to close

Ask: "I found N bugs. Want me to close them all?"

Context Awareness

Pronoun Resolution

When user says "it" or "that issue":

• If exactly one issue mentioned in last 3 messages → use it
• If multiple issues mentioned → ask: "Which issue - TES-123 or TES-456?"
• If no issue in last 5 messages → ask: "Which issue are you referring to?"

After CREATE:

User: "create a bug in TES" → TES-789 created
User: "assign it to me"
→ "it" = TES-789 (the issue just created)

User: "create a bug in DEMO" → DEMO-105 created
User: "show me the details of the bug we just created"
→ "the bug we just created" = DEMO-105 (use jira-issue to retrieve details)

After SEARCH:

User: "find all open bugs" → Found TES-100, TES-101, TES-102
User: "close them"
→ "them" = the search results (use jira-bulk)

Project Scope

When user mentions a project:

• Remember it for subsequent requests in this conversation
• "Create a bug in TES" → TES is now the active project
• "Create another bug" → Use TES implicitly
• Explicit project mention updates the active project

Context Expiration

After 5+ messages or 5+ minutes since last reference:

• Re-confirm rather than assume: "Do you mean TES-123 from earlier?"
• Don't guess when context is stale

Common Workflows

Create Epic with Stories

1. Use jira-agile to create the epic → Note epic key (e.g., TES-100)
2. Use jira-issue to create each story with --epic TES-100 flag to link during creation
   • Alternatively: create stories first, then use jira-as agile epic add-issues to link existing issues
3. Confirm: "Created epic TES-100 with N stories"

Bulk Close from Search

1. Use jira-search to find matching issues
2. Use jira-bulk with --dry-run to preview
3. Confirm count with user before executing

Data Passing Between Steps

When one skill's output feeds another:

• Capture entity IDs from responses (e.g., epic key from jira-agile)
• State this explicitly: "Created EPIC-123. Now creating stories..."
• Reference captured data in subsequent operations

Error Handling

If a skill fails:

• Report the error clearly
• Suggest recovery options from docs/SAFEGUARDS.md
• Offer alternative approaches

If a skill is not available:

• Acknowledge the limitation
• Suggest alternatives from the Quick Reference table

Permission Awareness

Before operations that might fail due to access:

• Check if user has mentioned permission issues before
• Suggest jira-admin for permission checks when blocked

Discoverability

• /jira-assistant-skills:browse-skills - List all skills with descriptions
• /jira-assistant-skills:skill-info <name> - Detailed skill information

If user asks "what can you do?" or similar:

• Show the Quick Reference table
• Offer to explain specific skills

What This Hub Does NOT Do

• Execute JIRA operations directly (always delegates)
• Guess when uncertain (asks instead)
• Perform destructive operations without confirmation
• Route to deprecated or unavailable skills without warning