The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the user to execute shell commands to run scripts (node scripts/generate_images.js and python3 scripts/publish_x.py). It also requires installing external dependencies via pip install playwright. While these are functional requirements, they represent a standard command execution surface.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided recruitment information (role name, responsibilities, etc.) to generate social media copy.
Ingestion points: User input for job details in SKILL.md workflows.
Boundary markers: None explicitly defined in the scripts to isolate user data from instructions.
Capability inventory: The script scripts/publish_x.py has the capability to automate browser actions, including filling text and clicking buttons on X.com.
Sanitization: There is no evidence of sanitization for the job descriptions before they are filled into the browser composer. An attacker could potentially inject instructions into the job description that the agent might follow if it were processing the output further.
[SAFE]: The browser automation in scripts/publish_x.py is designed to be interactive (headless=False), meaning the user remains in control of the login process and can inspect the post before it is finalized. This is a security best practice for automation involving social media accounts.

x-recruiter