The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes content from arbitrary external URLs.
Ingestion points: Content is fetched from user-provided URLs using curl, reader, or trafilatura as described in SKILL.md.
Boundary markers: Absent. The extracted content is displayed as a preview and saved to a file without clear delimiters or instructions to ignore embedded commands.
Capability inventory: The skill utilizes Bash for file operations and Write for saving content (as defined in SKILL.md).
Sanitization: While filenames are sanitized for filesystem compatibility, the extracted article text itself is not sanitized or filtered for malicious instructions.
[EXTERNAL_DOWNLOADS]: The skill suggests installing external dependencies to achieve its functionality.
Findings: Suggests installing reader-cli and @mozilla/readability-cli via npm, and trafilatura via pip. While Mozilla and Trafilatura are well-known, third-party packages like reader-cli should be manually verified before installation.
[COMMAND_EXECUTION]: The skill executes shell commands and Python scripts via Bash to handle data.
Findings: Uses python3 -c to run embedded scripts for HTML and JSON parsing. It employs variables like $ARTICLE_URL within shell commands. While it uses double quotes to mitigate simple shell injection, safety depends on the agent providing a valid URL.

article-extractor