Skip to content

PROMPT_CONFIGURATION.md

Latest commit

 

History

History
189 lines (143 loc) · 4.86 KB

PROMPT_CONFIGURATION.md

File metadata and controls

189 lines (143 loc) · 4.86 KB

Prompt Configuration Guide

Overview

The Canny MCP Server supports configurable prompts -- reusable templates that appear as /commands in Claude Code and other MCP clients. The server ships with 5 built-in prompts and lets you add your own.

Built-in Prompts

These 5 prompts work without configuration:

Prompt Arguments Purpose
weekly_triage boardID (opt), includeTrends (opt) Weekly feedback review
sprint_planning velocity (req), sprintLength (opt) Sprint backlog with RICE scoring
executive_summary month (opt) Monthly executive report
jira_sync_status none Jira-Canny sync health
customer_impact featureID (req) Feature impact analysis

Usage in Claude Code:

/weekly_triage
/sprint_planning velocity=21
/executive_summary month="January 2026"
/customer_impact featureID=68fac703a1c52202e2a1d917

Adding Custom Prompts

Edit config/default.json:

{
  "canny": { "..." : "..." },
  "server": { "..." : "..." },
  "logging": { "..." : "..." },
  "prompts": [
    {
      "name": "daily_standup",
      "description": "Prepare daily standup",
      "template": "Summarize yesterday's feedback:\n1. New posts\n2. Status changes\n3. Urgent items"
    }
  ]
}

Rebuild and restart:

npm run build

Then use in Claude Code:

/daily_standup

Prompt Format

{
  "name": "prompt_name",
  "description": "What this prompt does",
  "arguments": [
    {
      "name": "arg1",
      "description": "Description with example (e.g., 21)",
      "required": true
    }
  ],
  "template": "Prompt text with {arg1}"
}

Fields

Field Required Description
name Yes Unique identifier (lowercase, underscores)
description Yes User-facing description
arguments No List of argument definitions
template Yes Prompt text with {placeholder} syntax

Template syntax

  • {argumentName} inserts the argument value
  • \n creates a line break
  • Markdown formatting is supported (**bold**, numbered lists, etc.)
  • Undefined optional arguments become empty strings

Examples

No arguments

{
  "name": "daily_standup",
  "description": "Prepare daily standup",
  "template": "Summarize yesterday's feedback:\n1. New posts\n2. Status changes\n3. High-priority updates"
}

Required argument

{
  "name": "sprint_planning",
  "description": "Plan sprint backlog",
  "arguments": [
    {
      "name": "velocity",
      "description": "Sprint velocity in story points (e.g., 21)",
      "required": true
    }
  ],
  "template": "Based on velocity of {velocity} points:\n1. Find highest-priority posts\n2. Apply RICE scoring\n3. Suggest items fitting velocity"
}

Usage: /sprint_planning velocity=21

Multiple arguments

{
  "name": "competitive_analysis",
  "description": "Analyze competitor mentions in feedback",
  "arguments": [
    {
      "name": "competitor",
      "description": "Competitor name",
      "required": true
    },
    {
      "name": "timeframe",
      "description": "Time period (default: 30 days)",
      "required": false
    }
  ],
  "template": "Analyze requests mentioning {competitor} in {timeframe}:\n1. Most mentioned features\n2. Gaps in our product\n3. ARR impact\n4. Prioritization recommendation"
}

Usage: /competitive_analysis competitor=Salesforce timeframe="last 90 days"

How Prompts Load

  1. Server starts and loads config/default.json
  2. 5 built-in prompts load from src/prompts/index.ts
  3. Custom prompts load from config.prompts
  4. Both sets merge into the available prompts list

Check the server log for prompt counts:

Canny MCP Server initialized { prompts: 7, builtIn: 5, custom: 2 }

Best Practices

Naming -- Use descriptive lowercase names with underscores: weekly_triage, churn_risk_analysis. Avoid prompt1 or test.

Descriptions -- Explain the workflow in a short sentence: "Identify customers at risk of churning based on feedback signals."

Arguments -- Include examples in descriptions: "Sprint velocity in story points (e.g., 21)". Mark only truly required arguments as required.

Templates -- Structure with numbered sections. Be specific: "Analyze feedback from the last 7 days and identify the top 3 themes by vote count" beats "Analyze feedback."

Troubleshooting

Prompts not appearing

  1. Validate JSON: cat config/default.json | python3 -m json.tool
  2. Rebuild: npm run build
  3. Restart your MCP client
  4. Check server logs for the prompt count

Placeholders not replaced

Wrong: "Use ${velocity} points" Correct: "Use {velocity} points"

Use {name} syntax, not ${name}.

Arguments empty

  • Check that the argument name in the config matches the {placeholder} in the template exactly
  • Ensure you provide the argument when invoking the prompt