The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill contains a shell command that downloads a script from a remote URL and pipes it directly into the bash shell, which allows for arbitrary code execution without verification.
Evidence: curl -fsSL https://raw.githubusercontent.com/xdevplatform/xurl/main/install.sh | bash in SKILL.md.
[EXTERNAL_DOWNLOADS]: The skill references and encourages installation from several unverified external repositories managed by the 'xdevplatform' organization.
Evidence: NPM package @xdevplatform/xurl, Homebrew formula xdevplatform/tap/xurl, and Go module github.com/xdevplatform/xurl@latest.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from an external social media API (X API), which is an untrusted source that could deliver malicious instructions.
Ingestion points: xurl search, xurl read, xurl timeline, xurl mentions, and xurl dms commands in SKILL.md.
Boundary markers: While some safety warnings are present (e.g., instructions not to read ~/.xurl), there are no explicit delimiters or instructions to ignore embedded commands in the retrieved API data.
Capability inventory: The skill has the ability to execute shell commands, perform network requests, and potentially write to the filesystem via the xurl binary.
Sanitization: No evidence of sanitization, filtering, or validation is present for data ingested from the X API.
[COMMAND_EXECUTION]: The skill provides the agent with extensive capabilities to execute shell commands with user-controlled or API-controlled parameters.
Evidence: Support for raw endpoint access via xurl -X POST /2/tweets -d '{"text":"Hello world!"}' and various shortcuts for interacting with the X platform.

xurl