The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions for the agent to install third-party tools such as trafilatura (via pip) and @mozilla/readability-cli (via npm). While these are popular libraries, they represent a remote dependency risk if the package registries or specific versions are compromised.
[COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to execute shell commands, including piping content between curl, sed, tr, and python3. There is a risk of command injection if the URL or extracted article title (used for filenames) is not correctly handled, though the provided scripts include sanitization steps for filenames.
[PROMPT_INJECTION]: As a tool designed to scrape and process external web content, the skill is susceptible to indirect prompt injection. If an attacker controls the content of a URL being processed, they could embed instructions designed to manipulate the agent's behavior during the 'preview' or 'verification' steps.
[DATA_EXFILTRATION]: While not performing exfiltration itself, the skill uses curl to fetch content from user-provided URLs. This could be abused for internal network scanning (SSRF) if the agent is operating within a restricted environment.

article-extractor