By name: connector-observability description: Monitor, debug, and improve a published Claude connector using directory observability signals like adoption, error rates, and latency.
Connector observability
Use this skill to review the directory observability dashboard for a published connector, interpret the signals it provides, and turn them into concrete reliability/performance follow-ups.
Instructions
Confirm scope and access
- Confirm the connector is published in the Claude directory.
- Confirm you have access (Team or Enterprise with Admin/Owner, or an Enterprise custom role with the Libraries permission).
Review adoption signals
- Active users
- Total tool calls
- Directory rank over time
Review reliability and performance signals
- Health score
- Error rates
- Latency overview
- Per-tool error breakdown (use this to pinpoint which tool is failing)
Break down usage by product surface
- Compare tool calls across product surfaces (e.g., Claude, Claude Code, Cowork) to understand where usage is coming from.
Produce an action list
- Summarize the biggest issues/opportunities indicated by the dashboard.
- Propose the next steps (e.g., investigate top failing tools, address latency, validate changes by tracking error rate trends).
Examples
Example: quick review checklist
- Adoption: active users, total tool calls, directory rank trend.
- Quality: health score, error rate trend, latency overview.
- Debugging: per-tool error breakdown.
- Distribution: product breakdown (Claude vs Claude Code vs Cowork).