AI Agent Skills
AI agent skills are portable knowledge packages that give coding agents the context they need to work with Dynatrace. dtctl ships with its own skill and integrates with the broader Dynatrace for AI skill collection.
What are skills?
Skills follow the Agent Skills open format. They are small, structured files that teach AI coding agents about a specific domain or tool. Agents load only what they need: a short description for discovery, full instructions when relevant, and detailed reference files on demand.
Skills work with Claude Code, GitHub Copilot, Cursor, Kiro, Junie, OpenCode, OpenClaw, Gemini CLI, and many other compatible tools.
dtctl skill
dtctl includes a built-in skill at skills/dtctl/ that teaches AI agents how to operate dtctl: which commands to run, what flags to use, how to read output, and how to chain operations together.
Install with skills.sh
npx skills add dynatrace-oss/dtctl
Install with dtctl
dtctl skills install # Auto-detects your AI agent
dtctl skills install --for claude # Or specify explicitly
dtctl skills install --global # User-wide installation
dtctl skills install --cross-client # Cross-client (.agents/skills/)
dtctl skills status # Check installation status
Manual installation
Copy the skill directory into your agent’s skills path:
cp -r skills/dtctl ~/.agents/skills/ # Cross-client (any agent)
cp -r skills/dtctl ~/.github/skills/ # GitHub Copilot
cp -r skills/dtctl ~/.claude/skills/ # Claude Code
cp -r skills/dtctl ~/.openclaw/workspace/skills/dtctl/ # OpenClaw
Dynatrace domain skills
The dtctl skill teaches agents how to use the CLI tool. For deeper Dynatrace domain knowledge, install the skills from Dynatrace/dynatrace-for-ai:
npx skills add dynatrace/dynatrace-for-ai
These skills cover:
| Category | Skills |
|---|---|
| DQL & Query Language | DQL syntax rules, common pitfalls, query patterns |
| Observability | Services, frontends, distributed tracing, hosts, Kubernetes, AWS, logs, problems |
| Platform | Dashboards, notebooks |
| Migration | Classic entity-based DQL to Smartscape equivalents |
The domain skills provide context (how to write DQL queries, which metrics to use for service health, how to navigate distributed traces) while the dtctl skill provides the operational tool to act on it. Together they give AI agents everything they need to work with Dynatrace effectively.
You can install all skills without penalty. Agents use progressive disclosure and only load what they need for the current task.
Runtime bootstrapping
In addition to skills, AI agents can discover dtctl’s capabilities at runtime:
# Compact command catalog (ideal for system prompts)
dtctl commands --brief -o json
# Full catalog with descriptions and flag documentation
dtctl commands -o json
# Human-readable how-to guide
dtctl commands howto
See AI Agent Mode for details on structured JSON output, auto-detection, and the --agent flag.