The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill manages sensitive API keys for Phoenix and Arize AX. It implements proactive security measures to prevent credential leaking, specifically instructing the agent to never ask for keys in chat transcripts. Instead, it utilizes a local credential-entry helper that creates a secure, owner-only .env file (permissions 0o600), which is a security best practice.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from Phoenix (traces, datasets, and experiments). This presents a standard attack surface where malicious data could attempt to influence the agent. However, the skill provides explicit instructions to the agent to avoid placing raw data payloads into the conversation transcript, using bundled Python scripts for structural processing, and only reporting summarized counts. These boundaries effectively mitigate the risk of the agent obeying instructions embedded in the migrated data.
[REMOTE_CODE_EXECUTION]: The skill executes bundled scripts (scripts/migrate.py and scripts/migrate_data.py) within a virtual environment. All dependencies (arize, arize-phoenix-client, httpx, python-dotenv) are official vendor packages or standard libraries. The skill does not download or execute code from untrusted remote sources.
[COMMAND_EXECUTION]: The skill uses shell commands to manage its environment and run its helper scripts. Access is restricted to the local workspace and configured API endpoints. There is no evidence of arbitrary command injection vulnerability as user inputs (like project names) are handled via standard argument parsing in the scripts.
[SAFE]: The skill's behavior is consistent with its stated purpose of migrating data between Arize platforms. It uses secure file permissions, performs thorough verification of migrated data, and includes detailed logging for troubleshooting without exposing secrets.

arize-phoenix-migration