The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill enables an agent to process untrusted data from various external sources, creating a potential vector for indirect prompt injection. Ingestion points: Untrusted content can enter the agent's context through uploads screenshot (rendering web URLs or local HTML), uploads ingest (scanning GitHub pull request and issue comments), and uploads put --url (fetching remote assets). Boundary markers (absent): The documentation does not define specific delimiters or instructions for the agent to ignore potentially malicious embedded content. Capability inventory: The uploads tool can perform network egress, filesystem writes, and browser-based rendering. Sanitization (absent): While screenshots are rendered in a browser, the agent may still be influenced by visual or textual information extracted from the captures.
[EXTERNAL_DOWNLOADS]: The skill documents the use of uploads put --url, which permits the retrieval of content from arbitrary HTTPS endpoints. This is a functional requirement for the tool but constitutes a network ingestion point. Connections to uploads.sh and its subdomains are directed to the vendor's infrastructure.
[COMMAND_EXECUTION]: The skill provides detailed instructions for driving the uploads CLI, which involves the execution of shell commands that interact with the local environment, including file management and network communication.
[DYNAMIC_EXECUTION]: The uploads screenshot tool includes --eval and --init-script flags, allowing the execution of arbitrary JavaScript within the browser context during capture. This mechanism enables dynamic code execution on the target page.

uploads-cli