The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs users to install industry-standard libraries including the OpenTelemetry SDK (opentelemetry-sdk, @opentelemetry/sdk-trace-node) and the OpenAI client from official package registries (PyPI and NPM). These are well-known tools for monitoring and AI integration.
[INDIRECT_PROMPT_INJECTION]: The skill implements a standard AI agent loop that processes external data from user inputs and tool results. This creates a potential surface for indirect prompt injection where data could contain instructions meant to influence the model.
Ingestion points: The messages array processed in the run_agent function in both references/python.md and references/typescript.md.
Boundary markers: The templates do not currently include explicit delimiters or "ignore instructions" warnings in the system prompt examples.
Capability inventory: The skill scripts do not include high-risk capabilities such as file system modifications, arbitrary command execution, or network access beyond the defined tracing and AI provider endpoints.
Sanitization: The skill explicitly provides instructions in SKILL.md (Step 5) on how to redact sensitive content and handle large payloads before they are serialized and exported.
[SAFE]: The skill includes security-conscious guidance, specifically warning developers never to include API keys or authorization headers in trace attributes and advising against hardcoding secrets in client-side code.

maple-agent-tracing-opentelemetry