The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external dashboard configuration data (JSON) which represents a potential injection surface if the dashboard content (titles, descriptions) contains malicious instructions.
Ingestion points: Data is ingested through the signoz_get_dashboard tool response as described in SKILL.md.
Boundary markers: No specific delimiters are used for the interpolated dashboard content.
Capability inventory: The skill's instructions are limited to generating natural language explanations; it does not include commands for file writing, network operations, or code execution.
Sanitization: The instruction to "translate query intent into plain English" and "interpret" content serves as a semantic filter, treating the ingested data as information to be described rather than instructions to be followed.
[COMMAND_EXECUTION]: The skill references a CLI command signoz-mcp-setup. This is provided as troubleshooting advice for the user to initialize the required MCP server tools, rather than an instruction for the agent to execute commands autonomously or silently.

signoz-explaining-dashboards