The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The installation instructions provide a command to download and execute a script from 'https://raw.githubusercontent.com/xdevplatform/xurl/main/install.sh' by piping it directly to 'bash'. This execution method is inherently dangerous as it runs remote code with local user privileges.
[EXTERNAL_DOWNLOADS]: The skill requires downloading external assets from GitHub, npm (@xdevplatform/xurl), and Go package registries.
[COMMAND_EXECUTION]: The skill is primarily a wrapper for the 'xurl' CLI tool, allowing the agent to perform sensitive actions such as posting, deleting content, and sending direct messages on the user's behalf.
[CREDENTIALS_UNSAFE]: The skill manages and references the '~/.xurl' file, which stores sensitive OAuth tokens and API credentials. Accessing or interacting with this file path poses a risk of credential exposure.
[PROMPT_INJECTION]: The skill retrieves arbitrary content from X (tweets, DMs, search results), creating a surface for indirect prompt injection.
Ingestion points: Data is pulled from the X API through commands like 'xurl read' and 'xurl search'.
Boundary markers: The instructions do not define delimiters to separate untrusted API content from system instructions.
Capability inventory: The agent can perform impactful write operations (posting, messaging) based on the ingested data.
Sanitization: There is no documented mechanism to sanitize or validate the content fetched from the API before the agent processes it.

xurl