The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill uses elevated privileges via sudo to update package repositories and install the ffmpeg utility in the execution environment.
[COMMAND_EXECUTION]: The skill uses execSync to run shell commands, including a directory listing to find video files and an ffmpeg command to convert recordings. The conversion command incorporates the output path variable derived from command-line arguments without explicit sanitization, which may lead to command injection if malicious paths are provided.
[EXTERNAL_DOWNLOADS]: The skill downloads and installs the playwright package and its required Chromium browser binaries from the official npm registry.
[DYNAMIC_EXECUTION]: Browser interactions, such as clicking elements and filling forms, are dynamically executed at runtime based on a JSON-formatted list of actions parsed from user input.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill accepts an external URL and a JSON-encoded list of interaction steps through command-line arguments (found in SKILL.md).
Boundary markers: Absent; there are no clear delimiters or instructions to the agent to disregard potentially malicious commands embedded in the processed data.
Capability inventory: The skill has the ability to navigate to arbitrary URLs, interact with web page elements, write files to the local system, and execute shell commands.
Sanitization: While the skill parses the input JSON, it does not perform validation or sanitization on the selectors or values before passing them to the browser automation engine.

visual-evidence