The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the dtctl CLI tool for managing Dynatrace notebooks, involving commands for querying (dtctl query), fetching data (dtctl get notebook), and applying configurations (dtctl apply). The instructions proactively include security guidance to wrap DQL strings in single quotes to mitigate shell-level command injection risks.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection because it ingests and processes notebook JSON content (including markdown blocks and DQL queries) retrieved from external Dynatrace environments.
Ingestion points: Notebook content is fetched using dtctl get notebook <id> as outlined in references/analyzing.md.
Capability inventory: The agent is authorized to execute shell commands via dtctl and perform file system operations (writing and deleting notebook.json) as described in references/create-update.md.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the ingested notebook content as data rather than instructions.
Sanitization: While the skill mandates DQL syntax validation, it lacks specific instructions for sanitizing or escaping natural language instructions that might be embedded within the notebook's markdown sections.

dt-app-notebooks