| name | countly-prd |
| description | Countly production environment specialist for product analytics, event tracking, and crash reporting. Use when asked to: validate event instrumentation, analyze user sessions, check retention metrics, investigate crash reports, or debug analytics flows in the production environment. Trigger keywords: countly-prd. Do NOT use for production analytics — use countly-prd instead.
|
| compatibility | Requires countly-prd agent (countly-prd.agent.md) and countly-prd MCP server (HTTP). |
Countly Production Environment Skill
Agent Delegation
Delegate ALL operations to the countly-prd agent.
"Use the countly-prd agent to complete this task."
The countly-prd agent provisions the countly-prd MCP server and has the required tools
for sessions, events, retention, crashes, user flows, and performance analytics.
Description
Access Countly production environment for product analytics, user behavior tracking,
crash reporting, and application performance monitoring in production environment.
Capabilities
Sessions
- Analyze user session counts, durations, and frequency
- Segment sessions by device, OS, app version, or country
Events
- Track and validate custom event data (counts, segmentation, sums)
- Verify event properties and segmentation values are captured correctly
- Analyze funnel progression across event sequences
Retention
- Monitor day-N retention metrics for cohorts
- Compare retention across app versions or segments
Crashes & Errors
- Investigate crash reports and error traces from builds
- Group crashes by type, OS version, or app version
- Verify crash reporting instrumentation is working
User Flows
- Analyze user journeys and navigation flows in apps
- Identify drop-off points in onboarding or checkout flows
Performance
- Monitor application response times and slow requests
- Analyze network request performance from devices
Segmentation
- Segment any metric by demographics, device, OS, app version, or custom properties
Key Concepts
- App ID / App key: Each Countly application has a unique identifier — always confirm the target app before querying.
- Event: A named action tracked by the SDK (e.g.,
add_to_cart); events carry a count, sum, and optional segmentation properties.
- Segmentation: Key-value properties attached to events or sessions; used to filter and group analytics data.
- Retention: The percentage of users who return on day N after first use; measured per cohort.
- Production environment: Data comes from production environment and reflect real user behavior.
Workflow
- Connectivity probe — call a lightweight read-only tool (e.g., list apps or fetch server info) as the first action. If it fails, stop immediately and report: "countly-prd MCP server is unavailable. Cannot proceed."
- Identify app — confirm the target application name or ID from the user's request before constructing any query.
- Query — use the appropriate tool for sessions, events, retention, crashes, flows, or performance.
- Report using this structure:
- Query: what was analysed and for which app
- Time range: start → end
- Results summary: key metrics, counts, or patterns
- Anomalies / issues: unexpected values, missing events, error spikes
- Recommended actions: next steps or follow-up validations
Usage Examples
Validate Event Tracking
countly-prd Verify that the "add_to_cart" event is being tracked correctly in staging
Crash Debugging
countly-prd Show crash reports from the latest build
Performance Monitoring
countly-prd Analyze response times in the staging app
User Flow Validation
countly-prd Show user flows through onboarding in the production environment
Retention Check
countly-prd Show day-7 retention for the latest release
Configuration
MCP server is called countly-prd (HTTP type) and is configured in countly-prd.agent.md.
Environment variables are defined in .codex/.env.
Do NOT search the filesystem for mcp-config.json directly — always route through the countly-prd agent file.
Best Practices
- Always run a connectivity probe before executing any user-requested query
- Always confirm the target app before querying — wrong app ID silently returns unrelated data
- Never mix
countly-prd and countly-prd in the same request
Limitations
- Production environment reflect real user behavior or production volumes
- Different data retention policies than production
- Write or configuration operations require appropriate Countly permissions
Environment Isolation
CRITICAL: Never use countly-prd and countly-tst in the same request. Choose only one environment per query.