The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local CLI tools (agy, claude, jq, curl) to perform its primary function of model orchestration. It implements strong security practices by writing inputs to temporary files and using jq --rawfile to prevent shell injection, avoiding direct interpolation of untrusted data into command strings.
[EXTERNAL_DOWNLOADS]: The skill makes network requests to generativelanguage.googleapis.com and api.anthropic.com. These are official endpoints for well-known and trusted AI services (Google Gemini and Anthropic Claude), and these interactions are consistent with the skill's stated purpose.
[DATA_EXFILTRATION]: While the skill sends data to external model APIs, it does so within the scope of its core functionality. It utilizes isolation techniques, such as creating empty temporary directories and using env -i to strip sensitive environment variables from sub-processes, which reduces the risk of accidental local data leakage.
[CREDENTIALS_UNSAFE]: The skill follows standard safe practices for secret management by instructing users to set environment variables (ANTHROPIC_API_KEY, GEMINI_API_KEY) rather than hardcoding credentials. It includes checks to ensure these keys are present before attempting network operations.

advisor-orchestrator-worker