/create-prompt

Create prompts for sidecar workspaces. Covers prompt structure (name, ticketMode, body), template variables (ticket with fallbacks), config file locations (global vs project), and scope overrides. Use when creating or modifying prompts in sidecar config files.

marcus@marcus

GitHub

From plugin

sidecar

1k17 skills

Install

$ npx -y skills add marcus/sidecar --skill create-prompt --agent claude-code

How it fires

How this skill gets triggered: by you, by Claude, or both.

Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.
You can call itInvoke it directly when you want it.
Slash command/create-prompt

Context preview

The summary Claude sees to decide when to auto-load this skill.

Create prompts for sidecar workspaces. Covers prompt structure (name, ticketMode, body), template variables (ticket with fallbacks), config file locations (global vs project), and scope overrides. Use when creating or modifying prompts in sidecar config files.

SKILL.md

create-prompt.SKILL.md

name: create-prompt
description: Create prompts for sidecar workspaces. Covers prompt structure (name, ticketMode, body), template variables (ticket with fallbacks), config file locations (global vs project), and scope overrides. Use when creating or modifying prompts in sidecar config files.

Creating Prompts

Prompts are reusable templates that configure the initial context for agents when creating workspaces. They help standardize common workflows like code reviews, bug fixes, or feature implementation.

Config File Locations

| Scope | Path | Override Priority | |---------|---------------------------------------|-------------------| | Global | `~/.config/sidecar/config.json` | Lower | | Project | `.sidecar/config.json` (in project) | Higher |

Project prompts override global prompts with the same name (case-sensitive match).

Config Format

{
  "prompts": [
    {
      "name": "Code Review",
      "ticketMode": "optional",
      "body": "Do a detailed code review of {{ticket || 'open reviews'}}.\nFocus on correctness, edge cases, and test coverage."
    }
  ]
}

Prompt Fields

name (required)

Display name shown in the prompt picker. Keep it concise.

ticketMode (optional)

Controls how the ticket/task field behaves:

| Mode | Behavior | |------------|-----------------------------------------------| | `optional` | Task field shown, can be empty (default) | | `required` | Task must be selected before creating | | `none` | Task field is hidden, prompt stands alone |

body (required)

The prompt text sent to the agent. Supports template variables. Use `\n` for newlines.

Template Variables

`{{ticket}}`

Expands to the selected task ID. Returns empty string if no task selected.

"body": "Fix issue {{ticket}}."
// With task td-abc123: "Fix issue td-abc123."
// Without task: "Fix issue ."

`{{ticket || 'fallback'}}`

Expands to task ID, or the fallback text if no task selected.

"body": "Review {{ticket || 'all open items'}}."
// With task td-abc123: "Review td-abc123."
// Without task: "Review all open items."

**Only single quotes work for fallback text:**

`{{ticket || 'default'}}` -- works
`{{ticket || "default"}}` -- does NOT work

Examples

Task-Required Workflow

{
  "name": "Implement Task",
  "ticketMode": "required",
  "body": "Begin work on {{ticket}}. Use td to track progress.\nRead the task description carefully before starting."
}

Standalone Workflow

{
  "name": "Run Tests",
  "ticketMode": "none",
  "body": "Run the full test suite. Fix any failures.\nReport a summary when complete."
}

Flexible Workflow

{
  "name": "Code Review Session",
  "ticketMode": "optional",
  "body": "Start a review session for {{ticket || 'open reviews'}}.\nCreate td tasks for any issues found."
}

Backlog Refinement

{
  "name": "Backlog Refinement",
  "ticketMode": "none",
  "body": "Start a backlog refinement session. Use td to find all tasks in the backlog starting with the oldest. For each task:\n\n1. Determine relevance - has it been implemented, is it still needed? If not relevant, comment on the task and close it. If questionable, leave open with notes.\n\n2. Update tasks with out-of-date code references to reflect the current state of the codebase.\n\nUse sub-agents to analyze multiple tasks in parallel. Present a report of findings and changes when complete."
}