Tencent Cloud CAT Network Quality Analysis

Problem

In network operations and incident triage, engineers often need to determine whether a CAT probe task has errors, latency issues, or overall quality degradation, or whether a specific probe capture is required. Manual workflows can scatter evidence across consoles, mix up time windows, and route ambiguous requests incorrectly.

How It Works

The skill delegates analysis to CAT AI Console and enforces explicit parameters: --start-time and --end-time are required in millisecond timestamps; --task-id is required except for multi-task comparison. It routes by priority: multi-task comparison, pcap capture, error, overall, and performance. For standard analysis, it collects QueryText, task ID, and time range, defaulting to the last 3 hours when unspecified, then runs the analysis synchronously for up to 300 seconds. The script stays silent and emits one JSON object to stdout. On success, the agent reads the report field and sends it as the message body verbatim, without summaries, prefixes, suffixes, or attachments. On failure, it reports the categorized error when code=1 and does not auto-retry.

Boundaries and Caveats

Pcap reports must be sent as an independent message. If the request mentions pcap, capture, or similar capture keywords, the flow prioritizes direct pcap routing instead of running error or overall analysis first. Multi-task comparison omits --task-id, placing main_task_id and compare_task_ids in the query text, with a limit of one main task and five comparison tasks; incomplete task selections must be clarified before invoking analysis.