awesome-content-publisher
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill has a large attack surface as it is designed to ingest and interpret content from external web pages (DOM elements, console messages, network logs) and local Markdown files. These ingestion points in
SKILL.mdand various platform notes are combined with high-capability tools likebrowser_run_code_unsafe. To mitigate risks, the skill includes explicit boundary markers and instructions to the agent to treat web content strictly as data and never as instructions, and to report any text on a page that appears to address the agent directly. - [DYNAMIC_EXECUTION]: The instructions frequently require the agent to generate and execute dynamic JavaScript snippets using
browser_run_code_unsafe. This is documented inreferences/browser-interaction.mdand multiple platform-specific files to overcome UI automation challenges, such as coordinate-based clicking, custom file input handling, and interacting with rich-text editor instances (TipTap, ProseMirror, Quill). - [COMMAND_EXECUTION]: In
references/browser-interaction.md, the skill describes a workflow to bypass browser limitations by starting a local HTTP server on127.0.0.1. This allows the agent to provide files directly to composers using the File System Access API. This requires the agent to execute shell commands to manage the lifecycle of this temporary server.
Audit Metadata