The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill makes network requests to several well-known and official service providers. It interacts with Yandex API endpoints (api.direct.yandex.com, api-metrika.yandex.net, oauth.yandex.ru) for campaign management, Roistat (cloud.roistat.com) for business analytics, and YouGile (ru.yougile.com) for project tracking. These connections are integral to the skill's stated purpose of performance marketing operations.
[COMMAND_EXECUTION]: The script scripts/codex_cli_swarm_manual_review.py uses the subprocess module to run the codex exec command. This is part of a "swarm review" feature designed to process large batches of search query data in parallel by spawning subordinate agent instances.
[DATA_EXFILTRATION]: The skill handles sensitive authentication data, such as OAuth tokens and API keys, by loading them from local project files (e.g., oauth_token.json, .env) and environment variables. This data is transmitted exclusively to the official API endpoints of the supported services. There is no evidence of credential harvesting or unauthorized data transmission.
[PROMPT_INJECTION]: Multiple prompt templates are included (e.g., templates/codex_swarm_search_worker_prompt.md) that contain instructions for data analysis. These templates follow security best practices by employing delimiters and explicit instructions to ignore any commands that might be embedded within the untrusted search query or placement data being processed.

yandex-performance-ops