co guide — task recipes
Recipes for which co commands to run, in which order, for a given situation (for AI agents)
co guide prints task-oriented recipes: which co commands to run, in which order, and what to read in their output. Without a topic it prints every topic, so one read gives an AI agent the whole map. Output is plain text without color.
co guide # every topic
co guide incident # one topic
co guide --list # topic names onlyTopics
| Topic | Covers |
|---|---|
| start | Conventions: auth, workspace, output, non-interactive use |
| cr | Change requests: create, review, checks, merge |
| ci | CI runs: why a run failed, logs, rerun |
| issue | Issues: find, create, update |
| deploy | Argo CD: is it deployed, why is a sync stuck |
| incident | A service is failing or slow: errors, latency, which deploy, traces, logs |
| metrics | Kubernetes / container / custom metrics by label |
| dashboards | Read a dashboard's current values; export / import |
| alerts | Alert rules: backtest a threshold before enabling it |
| ingest | Sending telemetry: ingest tokens, linking a service to its repo |
Topics for features not available on the current host (for example stage-only commands) are hidden.
Wiring it into an agent
Add one line such as "run co guide and read it before you start" to the agent's system prompt or project instructions (CLAUDE.md, AGENTS.md, ...) so it works from the current commands and flags.