The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes local shell scripts to perform its primary functionality.
Evidence: It executes gh-upload-media.sh and gh-upload-dir.sh located within the plugin's internal scripts/ directory.
[EXTERNAL_DOWNLOADS]: The skill transmits media data to external GitHub repositories.
Evidence: Uploads are directed to the vendor-owned repository ArcBlock/loop-agent-assets and potentially to github.com/user-attachments/assets/ as a fallback mechanism.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface due to its handling of external files and directory structures.
Ingestion points: The skill accepts <dir> and <file> paths from the user to identify content for uploading.
Boundary markers: The instructions do not define explicit delimiters or warnings to prevent the agent from being influenced by metadata or content within the processed files.
Capability inventory: The skill uses bash to run scripts that have full read access to the local filesystem and network access via the gh CLI and git protocols.
Sanitization: Basic file extension validation is performed to ensure only media types (e.g., .png, .jpg, .webm) are processed.

media-upload