The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of the opentelemetry-mcp package from PyPI and provides instructions for cloning its source code from a GitHub repository maintained by Traceloop. As these are established and well-known repositories for observability tools, the downloads are considered a standard part of the skill's operation.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection because it retrieves and interprets data from external observability backends. This data, which includes span tags and log messages, can be populated by users or external systems interacting with the monitored application.
Ingestion points: Data enters the agent's context through tools like get_trace, search_traces, and search_spans which fetch trace information from external APIs.
Boundary markers: There are no specific delimiters or instructions defined in the skill metadata to warn the agent about potentially malicious instructions embedded within the trace data.
Capability inventory: The skill allows the agent to analyze, search, and aggregate data to debug errors and monitor LLM performance, meaning injected instructions could influence these analytical tasks.
Sanitization: No sanitization or filtering logic is mentioned for the content retrieved from the trace backends before it is processed by the AI.

opentelemetry-mcp-server