setting-up-tasks - SKILL.md Agent Skill

name: setting-up-tasks description: > Guides initial setup of the waggle plugin — detects or configures MCP connections and runs provider-specific database initialization. Use this skill whenever the user wants to set up, initialize, or configure waggle for the first time, or connect a new data source provider. Triggers on: "setup waggle", "initialize task management", "configure notion tasks", "configure data source", "get started with waggle", "first-time setup", "connect provider". user-invocable: true

Waggle — Setup Guide

You are guiding the user through the initial setup of the waggle plugin.

Output Discipline

This skill runs as a multi-step pipeline, but the user only needs its outcomes. Do not narrate step transitions ("Now I'll...", "X done, next Y") and do not relay protocol internals — provider detection, config/schema checks, cache state, validation plumbing, view-server pushes. Surfacing them buries what actually matters.

Emit user-facing text only when it changes something for the user:

a prompt or confirmation that needs their input
an error or a warning
an intermediate result that changes the outcome (e.g., a non-PASS quality verdict and the gaps behind it — it explains why a task lands at a different status than expected)
the final result summary

Step 1: Check for Existing MCP Configuration

Inspect available MCP tools to detect any already-configured providers:

notion-* tools present → Notion MCP is already configured
SQLite/database tools present → SQLite is already configured

If a single provider MCP is already present

Use AskUserQuestion to confirm:

"I detected an existing [provider] MCP connection. Would you like to set up waggle using [provider]?"

If yes, check if the corresponding provider plugin is installed. This skill cannot delegate to detecting-provider because it runs before a provider exists (detecting-provider would error with "No waggle provider plugin found" and stop the setup flow).

First, decide which discovery source to inspect by detecting Cowork via the same multi-signal OR used by detecting-provider — Cowork iff any of:

The active system prompt mentions "Cowork" (e.g. an <application_details> block stating "Claude is powering Cowork mode")
A tool whose name matches mcp__cowork__* or mcp__cowork-onboarding__* is available
echo "$CLAUDE_CODE_IS_COWORK" returns "1" (legacy hint; absence is not evidence against Cowork because Bash subshells in Cowork are sandboxed)

Then branch:

If Cowork → look in <available_skills> for {provider}-provider
Otherwise → read ~/.claude/plugins/installed_plugins.json for waggle-{provider}@*

If not installed → error:

"The waggle-{provider} provider plugin is not installed. Install it first, then run setup again."

If installed, skip to Step 3 with that provider.

If multiple provider MCPs are present

Use AskUserQuestion to ask which one to use:

"I detected multiple data source MCPs: [list providers]. Which one should I set up waggle for?"

Then check if the corresponding provider plugin is installed (same discovery method as above). If not installed → error with install instructions. If installed, skip to Step 3 with the selected provider.

If no provider MCP is present

Continue to Step 2 to guide the user through MCP setup.

Step 2: Choose a Data Source and Configure MCP

Use AskUserQuestion to ask which data source the user wants to use:

"Which data source would you like to use for waggle?

SQLite — instant local setup, zero external dependencies

Notion — team collaboration via Notion workspace

Turso — remote SQLite for multi-agent sync (requires Turso account)"

Then guide MCP setup based on the environment:

Determining the Environment

Terminal CLI: ~/.claude/settings.json exists or CLAUDE_PLUGIN_ROOT is set
Claude Desktop: Global Instructions / CLAUDE.md is accessible from the current context

Claude Code — MCP Setup Instructions

Add the following to ~/.claude/settings.json under "mcpServers":

Notion:

"notion": {
  "type": "http",
  "url": "https://mcp.notion.com/mcp"
}

After adding, authenticate by visiting https://mcp.notion.com/mcp in a browser and following the OAuth flow. Then restart Claude Code and run the setup skill again.

Claude Desktop — MCP Setup Instructions

Notion: Open Claude Desktop settings → MCP Servers → Add Server → Enter https://mcp.notion.com/mcp. Authenticate with your Notion account when prompted. Then run the setup skill again.

SQLite — No MCP Required

SQLite requires no external MCP server. Proceed directly to Step 3.

Step 3: Run Provider-Specific Setup

The provider plugin's setup skill handles all database creation and configuration.

If the provider plugin is installed, instruct the user:

"Run 'setup {provider}' to initialize the {provider} provider."

The provider plugin's {provider}-setup skill is user-invocable and handles everything:

Database/schema creation
Config page creation (Notion) or config file generation
Validation and testing

Step 3.5: Custom Intake Source Configuration

Configure additional sources for the message intake skill. This step runs for all environments.

Prompt User

Use AskUserQuestion (multiSelect) to ask:

"The message intake skill reads from Slack/Teams/Discord by default. Would you like to also ingest from additional sources?"

"Additional messaging tools (e.g., Google Chat, Discord)"

"Spreadsheet to scan for new tasks"

"Other task management system (e.g., Jira, Linear)"

"Custom processing step (free-text)"

If the user selects none or says "skip", proceed to Step 4.

Collect Details

For each selected option, ask follow-up questions using AskUserQuestion:

Messaging tools → "Which tool? How should it be accessed (MCP tool name, API, etc.)?"
Spreadsheet → "Which spreadsheet? (name or URL) Which columns/rows indicate new tasks?"
Task system → "Which system? What filter criteria? How to access (MCP tools, API)?"
Custom step → "Describe what should happen and when (before/after standard intake):"

Compose the answers into natural language instructions that the ingesting-messages skill will follow at runtime.

Example:

# Custom Intake Sources

## Google Chat
After standard messaging intake, check Google Chat DMs for the past 24 hours using google-chat MCP tools.
Create tasks from actionable messages using the same classification rules.

## Sprint Tracker Spreadsheet
Scan the "Sprint Tracker" Google Sheet (https://docs.google.com/...) for rows added in the past 24 hours.
Create each new row as a Backlog task with the row contents as the description.

Save Instructions

Determine the environment (reuse execution_environment from detecting-provider):

CLI / Claude Desktop

Write the instructions to ~/.waggle/intake-prompt.md:

Confirm to user:

"Custom intake sources saved to ~/.waggle/intake-prompt.md. The ingesting-messages skill will read this file each time it runs. You can edit the file directly to update your configuration."

Cowork

Since Cowork has an ephemeral filesystem, output the instructions as a block the user must paste into their Global Instructions:

"Cowork does not have a persistent filesystem. Please add the following to your Global Instructions so it's available in every session:"
<waggle-custom-intake>
{composed instructions}
</waggle-custom-intake>

Step 3.6: Custom Task-Creation Rules

Collect business-logic rules for how new tasks should be created — for example, tag naming conventions, tag assignment per category, default priority rules, or how Acceptance Criteria should be phrased. These rules are applied by managing-tasks, ingesting-messages, and planning-tasks whenever they create or plan a task.

This is independent of Step 3.5 (custom intake sources). A user may configure one, both, or neither.

Prompt User

Use AskUserQuestion:

"Do you want to define project-specific rules for how waggle creates tasks? Examples: required tags, tag operation rules, default priorities, AC phrasing style. You can skip this and configure it later by editing ~/.waggle/task-creation-prompt.md."

If the user skips, proceed to Step 4.

Collect Rules

Ask follow-up questions via AskUserQuestion to capture:

Tag rules: required tags, naming conventions, category → tag mapping
Priority rules: when to default to Urgent / High / Medium / Low
Assignee defaults: any rules for routing tasks to specific members
AC / Execution Plan style: e.g. "AC must use Given/When/Then"

Compose the answers into natural language instructions.

Example:

# Custom Task-Creation Rules

## Tags
- All tasks in the `frontend/` directory must include the `frontend` tag.
- Security-related tasks (mentioning CVE, vulnerability, auth bypass) must include `security`.

## Priority
- Tasks mentioning production incidents or customer escalations default to Urgent.
- Tasks blocked on external vendors default to Low.

## Acceptance Criteria style
- Use Given/When/Then format for all user-facing feature tasks.
- Always include at least one testable criterion with a command or metric.

Save Rules

Determine the environment (reuse execution_environment from detecting-provider):

CLI / Claude Desktop

Write the rules to ~/.waggle/task-creation-prompt.md. Keep the file under 10 KB — the loader rejects anything larger. Do not paste text from untrusted sources into this file; it is concatenated into agent prompts and carries prompt-injection risk.

Confirm to user:

"Custom task-creation rules saved to ~/.waggle/task-creation-prompt.md. The managing-tasks, ingesting-messages, and planning-tasks skills will read this file each time they run. You can edit the file directly to update your rules. Security note: this file is trusted input — only put rules you authored yourself, and keep the file under 10 KB."

Cowork

Since Cowork has an ephemeral filesystem, output the rules as a block the user must paste into their Global Instructions:

"Cowork does not have a persistent filesystem. Please add the following to your Global Instructions so it's available in every session. Only paste rules you authored yourself — this text is concatenated directly into agent prompts."
<waggle-custom-task-creation>
{composed rules}
</waggle-custom-task-creation>