The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download and install the x-cli tool directly from a third-party GitHub repository (https://github.com/Infatoshi/x-cli.git) using the uv package manager.
[REMOTE_CODE_EXECUTION]: The installation command uv tool install git+https://github.com/Infatoshi/x-cli.git fetches and installs executable code from an external, unverified source onto the local system.
[COMMAND_EXECUTION]: The skill is designed to execute various shell commands using the installed x-cli tool. These commands often include arguments derived from user input or external data (e.g., tweet content, search queries), which are passed directly to the shell.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks as it ingests untrusted data from the X API (such as tweet content, mentions, and search results) and provides the agent with capabilities to perform actions (like posting or deleting tweets) based on that data.
Ingestion points: x-cli tweet search, x-cli user timeline, x-cli me mentions, and x-cli me bookmarks (SKILL.md).
Boundary markers: None explicitly defined in the provided instructions to isolate external data from instructions.
Capability inventory: x-cli tweet post, x-cli tweet reply, x-cli tweet delete, x-cli like, and x-cli retweet (SKILL.md).
Sanitization: The instructions recommend user confirmation for write actions, but do not specify technical sanitization or validation of the ingested content.

xitter