cx-olly

cx-olly is a skill for driving Coralogix's Observability Agent ("Olly") from the `cx` command-line tool. It lets an agent or a human ask natural-language questions about their observability data — alerts, metrics, logs — and receive AI-generated analysis, charts, and reports without hand-writing queries. The problem it solves is turning conversational, exploratory observability investigations ("Why is checkout latency high?", "What changed in the last hour?") into repeatable CLI calls that can be scripted, backgrounded, and chained.

The skill documents three commands: `cx olly ask` to send a message, and `cx olly artifacts list` / `cx olly artifacts get` to retrieve generated content such as charts and query results. It covers chat continuation via `--chat-id`, model selection (gpt-5.2 default, Claude Sonnet/Haiku, and others), timeout tuning, and an `--agent-to-agent-mode` flag that agents should pass for shorter sub-agent-style replies. Because investigations can take minutes, it recommends running `ask` asynchronously as a background process and polling for completion. Machine-readable output (`-o json`/`-o toon`) supports piping to `jq`, and artifact retrieval automatically decompresses gzip, parses JSON, and spills large text content to temp files.

It targets SREs, platform engineers, and LLM agents already using Coralogix who want conversational observability from the terminal or from automated workflows. Typical use cases include incident root-cause analysis, alert triage, ad-hoc log/metric questions, and scripted extraction of AI-generated charts. It is single-profile only and relies on the underlying `cx` CLI and Coralogix account for authentication.