The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a custom OpenTelemetry SpanProcessor that parses genkit:input and genkit:output span attributes at runtime to extract conversation transcripts and tool usage.
Ingestion points: The processor reads from genkit:input and genkit:output attributes in genkit-for-maple.ts, which contain raw data from agent interactions.
Boundary markers: No specific boundary markers or 'ignore' instructions are applied to the data being processed, as the processor expects structured JSON content generated by the Genkit framework.
Capability inventory: The code performs JSON.parse() on these attributes and re-maps the contents to gen_ai.* attributes, which are then transmitted over the network via the OTLP exporter.
Sanitization: The implementation does not sanitize the contents of the messages, reasoning, or tool results before mapping them to the new attributes, as it is designed for high-fidelity tracing.
[COMMAND_EXECUTION]: The skill provides standard npm installation commands for OpenTelemetry and Genkit dependencies. These are standard packages required for the stated functionality.
Evidence: npm install genkit @opentelemetry/sdk-node @opentelemetry/sdk-trace-base @opentelemetry/exporter-trace-otlp-proto in Step 2.
[DATA_EXPOSURE]: The skill is explicitly designed to export telemetry data (including model prompts, tool arguments, and results) to the Maple ingestion endpoints (ingest.maple.dev or ingest.eu.maple.dev). The documentation includes warnings to ensure users do not hardcode private keys and suggests using standard environment variable practices for handling the Authorization header.

maple-agent-tracing-genkit