The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to use npx @arizeai/phoenix-cli to download and run the Phoenix CLI. This is a standard deployment method for the official tool provided by the vendor.
[COMMAND_EXECUTION]: The skill documentation provides extensive shell command examples for managing LLM traces, spans, and profiles. These include piped operations with jq for data processing and shell while loops for bulk annotation of traces.
[PROMPT_INJECTION]: The skill creates an indirect prompt injection surface by directing the agent to process and analyze untrusted trace data.
Ingestion points: The agent fetches raw LLM inputs, outputs, and retrieved context from external sources using px trace get and px span list as described in SKILL.md and the open-coding workflow.
Boundary markers: No delimiters or safety instructions are defined to help the agent distinguish between trace data and instructions.
Capability inventory: The agent has the capability to execute shell commands and modify remote data (via annotate and add-note commands) based on its analysis of the untrusted trace content.
Sanitization: The instructions do not provide any mechanisms for sanitizing or validating trace content before it is processed by the agent.

phoenix-cli