The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run within scripts/metric_alert.py to execute ae-cli commands for ad-hoc SQL analysis. This is a primary function of the skill, designed to interface with the platform's CLI tool to fetch metric data.
Evidence: The script scripts/metric_alert.py builds command arrays (e.g., cmd = [ae_cli, "analysis", "adhoc", "run", ...]) and executes them using subprocess.run(cmd, capture_output=True, text=True, timeout=QUERY_TIMEOUT_SECONDS, check=False).
Mitigation: The script implements several safety checks, such as enforcing LIMIT clauses and requiring $part_date partition predicates to prevent expensive full-table scans. Additionally, it uses the platform's ad-hoc analysis API, which is typically read-only, limiting the impact of the executed SQL.
[SAFE]: All external downloads and tool invocations refer to the platform's own infrastructure (the ae-cli tool) or the user's configured notification webhooks (WeCom, DingTalk, Feishu, Slack). The script scripts/metric_alert.py is loaded from the skill's own assets via ae-cli agent +read-skill-script.
Evidence: SKILL.md and references/agent-setup.md define standard procedures for reading internal script assets and using the verified ae-cli tool (version 6.1.5). No untrusted third-party repositories or external script sources were identified.

metric-alert