The Agent Skills Directory

[CREDENTIALS_UNSAFE]: Potential credential exposure in system process list. The workflow.py script constructs a curl command that includes the Sorftime API key in the URL. By executing this via subprocess.run(), the secret key becomes visible to other users or processes monitoring the system's process table.
[COMMAND_EXECUTION]: Dependency on external command execution. The skill uses subprocess.run() to invoke curl for API interactions. While legitimate for its purpose, this increases the attack surface compared to using native Python HTTP clients like requests.
[PROMPT_INJECTION]: Indirect injection vulnerability surface. The skill processes external data (Amazon product titles, brand names) and renders it into generated Markdown and HTML reports without proper sanitization.
[PROMPT_INJECTION]: Ingestion points: Data fetched from the category_report tool in scripts/workflow.py and scripts/analyze_category.py.
[PROMPT_INJECTION]: Boundary markers: None. The skill does not use delimiters to wrap untrusted data before presentation to the user or agent.
[PROMPT_INJECTION]: Capability inventory: The skill can write files, execute subprocesses, and make network requests.
[PROMPT_INJECTION]: Sanitization: Report generation templates (e.g., assets/dashboard_template.html) use innerHTML to render product data, which is vulnerable to Cross-Site Scripting (XSS) if the source data contains malicious HTML.

category-selection