The Agent Skills Directory

[COMMAND_EXECUTION]: The skill interpolates user-supplied data (post_url) and external tool output (thumbnail_url) directly into shell commands for yt-dlp and curl. This creates a significant risk of command injection if the agent does not properly escape these variables (e.g., a URL containing ; command_here could trigger arbitrary execution).
[REMOTE_CODE_EXECUTION]: The skill uses uvx to dynamically download and execute the yt-dlp package from the Python Package Index (PyPI) at runtime. This involves fetching and running external code that is not pinned to a specific version or hash.
[CREDENTIALS_UNSAFE]: Instructions direct the agent to retrieve secrets from the macOS keychain using security find-generic-password. This exposes sensitive credentials to the shell environment and command-line history.
[DATA_EXFILTRATION]: Credentials retrieved from the local keychain are transmitted via curl to the Imgur API. While this is the intended functionality, the pattern of reading local secrets and sending them to a remote endpoint is a primary vector for exfiltration.
[EXTERNAL_DOWNLOADS]: The skill fetches content from arbitrary social media URLs and then uploads that data to api.imgur.com. This involves multi-step interactions with external, untrusted network resources.

social-to-imgur