Instance Analyze

Instance Analysis Skill

This skill provides guidance for analyzing a Canvas Medical instance's configuration before building plugins. The analysis informs plugin design decisions and identifies existing resources to leverage.

Execution standard: Run Python scripts and Python-based tooling with uv run ... (for scripts, uv run python <script>.py ...). Do not invoke bare python or pip.

When to Use This Skill

Use this skill when:

• Starting a new plugin project and need to understand the target environment
• Need to know what teams, roles, or questionnaires exist for plugin integration
• Want to document current instance configuration for reference
• Checking what plugins are already installed

Workflow

Step 1: Understand Plugin Intent

First, ask the user about the plugin they intend to build. This helps identify if there's already a related plugin and tailors the analysis to their specific needs.

{
  "questions": [
    {
      "question": "What plugin are you planning to build? Please describe its purpose briefly.",
      "header": "Plugin Intent",
      "options": [
        {"label": "Clinical Alert", "description": "Alerts for vitals, lab results, or clinical conditions"},
        {"label": "Workflow Automation", "description": "Automate tasks, appointments, or care coordination"},
        {"label": "Data Integration", "description": "Sync data with external systems or EHRs"},
        {"label": "Patient Engagement", "description": "Patient messaging, reminders, or education"}
      ],
      "multiSelect": false
    }
  ]
}

Store the user's response to use later when:

• Checking for existing plugins that might overlap or conflict
• Highlighting relevant teams, questionnaires, and note types in the report
• Making plugin development recommendations

Step 2: Discover Available Instances

Read credentials from ~/.canvas/credentials.ini to discover configured instances:

cat ~/.canvas/credentials.ini

Parse the file to extract all section names (instance names). Each section represents a configured Canvas instance:

[instance-name]
client_id=xxx
client_secret=yyy
root_password=zzz

Step 3: Get Instance Information

Ask the user which instance to analyze using AskUserQuestion, dynamically populating the options with instances discovered in Step 2:

{
  "questions": [
    {
      "question": "What Canvas instance should I analyze?",
      "header": "Instance",
      "options": [
        {"label": "<instance-1>", "description": "From credentials.ini"},
        {"label": "<instance-2>", "description": "From credentials.ini"},
        {"label": "...", "description": "Add all discovered instances"}
      ],
      "multiSelect": false
    }
  ]
}

If the user provides an instance name with the command (e.g., /analyze-instance plugin-testing), use that directly without asking.

After the user selects an instance, verify it has root_password configured. If missing, tell the user:

Please add root_password to your ~/.canvas/credentials.ini under the [{instance}] section:

[plugin-testing]
client_id=your_client_id
client_secret=your_client_secret
root_password=your_admin_password

Step 4: Fetch Configuration

Use the scraper script to extract configuration from the admin portal:

uv run python ${CLAUDE_PLUGIN_ROOT}/scripts/scrape_canvas_instance.py <instance_name> <root_password>

This script uses browser automation to:

1. Login to Admin Portal

   • Navigate to https://{hostname}.canvasmedical.com/admin/
   • Username: root
   • Password: root_password from credentials.ini

2. Extract Configuration from these pages:

   • /admin/api/careteamrole/ - Roles (where is active is true)
   • /admin/api/team/ - Teams
   • /admin/api/questionnaire/ - Questionnaires (where is active is true)
   • /admin/api/notetype/ - Note types (where is active is true)
   • /admin/api/notetype/ - Appointment types (where is active and is schedulable are true)
   • /admin/plugin_io/plugin/ - Installed plugins

Step 5: Generate Report

Create a comprehensive markdown report with this structure:

# Canvas Instance Configuration Report

**Instance**: {hostname}.canvasmedical.com
**Generated**: {date}

## Summary

| Category | Count |
|----------|-------|
| Roles | X |
| Teams | X |
| Questionnaires | X |
| Note Types | X |
| Appointment Types | X |
| Installed Plugins | X |

## Roles

| Name | Description |
|------|-------------|
| ... | ... |

## Teams

| Name | Members | Description |
|------|---------|-------------|
| ... | ... | ... |

## Questionnaires

| Name | Code | Status |
|------|------|--------|
| ... | ... | Active/Inactive |

## Note Types

| Name | Category |
|------|----------|
| ... | ... |

## Appointment Types

| Name | Duration | Category |
|------|----------|----------|
| ... | ... | ... |

## Installed Plugins

| Name | Version | Status |
|------|---------|--------|
| ... | ... | Active/Inactive |

## Plugin Development Recommendations

Based on this configuration:

- **Available Teams for Task Assignment**: [list teams that could receive tasks]
- **Relevant Questionnaires**: [questionnaires that might be relevant to the use case]
- **Existing Plugins**: [any plugins that might conflict or complement new development]

Step 6: Save Report

Save the report to instance-config-{hostname}.md in the current working directory.

Tell the user the file path and offer to explain any section in detail.

Integration with Plugin Spec

If {workspace_dir}/.cpa-workflow-artifacts/plugin-spec.md exists (where workspace_dir is the git repository root), read it and tailor the report to highlight:

• Teams that match the spec's task assignment needs
• Questionnaires relevant to the spec's data requirements
• Existing plugins that might overlap

Important Notes

1. Credentials Security: Never log or store credentials. Use them only for the session.

2. Read-Only: This analysis only reads configuration. It does not modify anything.

3. Relevance Filtering: When presenting the report, highlight items most relevant to the plugin being developed (if known from a previous brainstorming session).

4. Plugin Conflicts: Flag any installed plugins that might interact with what's being built.

Example Interaction

User: "I need to analyze the plugin-testing instance before building a vitals alert plugin"

You: [Uses AskUserQuestion to confirm instance]

User: Confirms plugin-testing

You: [Reads credentials and runs scraper script]

You: """I've analyzed the plugin-testing instance and saved the report to instance-config-plugin-testing.md.

Key findings relevant to your vitals alert plugin:

• 3 clinical teams available for task assignment: Nursing, Care Coordination, Providers
• No existing vitals-related plugins installed - no conflicts expected
• 12 active questionnaires - none specifically for vitals alerts

Would you like me to explain any section in more detail?"""

User: Yes / No