Create ticket

Create Ticket

Use this workflow when the user asks to create a ticket (for example: "create ticket", "open ticket", "log request").

This skill reads pm_tool from config to determine which PM integration to use. Currently supported: linear. The instructions below use Linear-specific MCP tools; when support for other PM tools is added, the tool calls will change but the workflow structure remains the same.

Config dependency

Read <workspace>/.cursor/skills-config.yaml before acting. If the file is missing, tell the user to run bootstrap-config first.

Required config keys:

• pm_tool — which PM tool to use (default: linear)
• linear.team_name, linear.team_id
• linear.default_project, linear.default_project_id
• linear.projects[]
• linear.labels[]
• linear.default_priority
• linear.custom_input

Guardrails

1. If any required data is unclear, ask the user before creating the issue.
2. Keep the final issue description structured for both humans and AI agents.

Cached identifiers

Use IDs from skills-config.yaml first to avoid repeated lookups. The config contains team ID, project IDs, and label IDs pre-resolved by bootstrap-config.

Fallback cache refresh

If Linear rejects a cached ID (not found/invalid):

1. Refresh only that entity:
   • team -> list_teams
   • project -> list_projects
   • label -> list_issue_labels
2. Update in-memory mapping for this run.
3. Continue without re-fetching unrelated IDs.

Required metadata policy

Team is always fixed to the value in config. If the user does not provide a field, ask whether to apply the default:

Field	Default	Prompt style
Project	linear.default_project	Choose from list
Priority	linear.default_priority	Apply default? Y/N
Label	—	Choose from list + None
Assignee	—	me / Unassigned / Other
Cycle	—	Current / None

Incorporate any additional rules from linear.custom_input.

Intake flow

1. Parse provided inputs from user message: title, optional body/context, optional metadata.
2. Ask for missing fields using structured multiple-choice questions.
3. For each missing field, ask "Apply default?" when a default exists.
4. For label, require one explicit choice from the config label list including None.
5. For assignee, require one explicit choice: me, Unassigned, or Other.
6. For cycle, require explicit choice: Current or None.
7. Resolve names to IDs using config cached mappings.

Ticket content standard (human + AI)

When creating a ticket description, use this template:

## Context
<background and why this is needed>

## Problem statement
<current issue/gap>

## Requested change
<what needs to be implemented>

## Acceptance criteria
- <criterion 1>
- <criterion 2>

## Validation / test plan
- <how to verify behavior>

## Risks / scope guardrails
- <known risk or constraint>

## Production verification notes
<prod checks required later, or "Not required">

If user gives loose notes, normalize them into the template before creating the issue.

Issue creation call

Use save_issue with:

• team: from linear.team_name
• title: concise, action-oriented
• description: structured markdown template above
• optional metadata from intake:
  • project
  • priority
  • labels (single-item array when label is selected; omit when None)
  • assignee
  • cycle (set current cycle only when user selects Current; omit when None)

Completion output format

After creating the issue, return:

1. Issue ID and title
2. URL
3. Applied metadata: team, project, priority, label, assignee, cycle
4. Defaults audit: fields defaulted vs. user-provided

Use this response shape:

Created:

- Issue: <ID> — <Title>
- URL: <url>

Applied metadata:
- Team: <value>
- Project: <value>
- Priority: <value>
- Label: <value>
- Assignee: <value>
- Cycle: <value or None>

Defaults used:
- <field>: <default used>

Provided by user:
- <field>: <value>