The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill identifies an attack surface where natural language input from a user (e.g., "calculate the error rate") is translated into PromQL expressions. Because these expressions are subsequently used in shell commands, malicious instructions within the user input could theoretically influence the execution flow.
Ingestion points: Natural language user requests described in the SKILL.md frontmatter and description.
Boundary markers: No explicit markers or instructions are provided to delimit user-supplied data from the generated query logic.
Capability inventory: The skill utilizes shell command execution via curl and jq to interact with remote Prometheus endpoints in SKILL.md workflows.
Sanitization: There is no evidence of input validation or sanitization logic to ensure the generated PromQL is safe before being interpolated into shell commands.
[DYNAMIC_EXECUTION]: The skill's documented workflows involve generating and executing shell commands at runtime. Specifically, SKILL.md provides examples of assembling a query expression into a variable (EXPR) and interpolating it directly into a curl command line. While standard for this use case, it represents a pattern of dynamic script generation from templates.

promql