The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-defined comments in source code files to drive its code generation logic, creating a surface where malicious instructions could be embedded in the markers to influence the output code.
Ingestion points: Source code files containing @otel: or @otel(observable): comment markers.
Boundary markers: The skill relies on specific comment prefixes but lacks formal delimiters to distinguish between the marker data and potentially malicious instructions.
Capability inventory: The skill is capable of modifying source code files, creating new metrics, and altering configuration settings.
Sanitization: The skill implements a two-phase technical design process and a review checklist to mitigate incorrect code generation.
[DATA_EXFILTRATION]: The skill is designed to read and modify sensitive project configuration files and internal documentation.
Evidence: The skill description and responsibilities explicitly involve editing the telemetry.metrics configuration in kibana.yml and reading internal README files from the src/platform/packages/private/opentelemetry/kbn-metrics/ directory.

kibana-otel-instrumentation