The Agent Skills Directory

Data Exposure (SAFE): The seo_checker.py script reads local project files to identify missing SEO tags. It does not access sensitive user directories (e.g., .ssh, .aws) or environment files. No network exfiltration logic (like requests or curl) is present.
Remote Code Execution (SAFE): The script uses only Python standard libraries (sys, json, re, pathlib). It does not use eval(), exec(), or subprocess to execute commands or external scripts.
Indirect Prompt Injection (LOW):
Ingestion points: scripts/seo_checker.py reads the content of project files provided via command-line arguments.
Boundary markers: No explicit delimiters are used to wrap the untrusted file content.
Capability inventory: The script has zero high-risk capabilities; it does not write to the filesystem, perform network requests, or execute shell commands.
Sanitization: Content is processed via regex and string matching to count tags. Since the tool only outputs diagnostic text to the console, the risk of an injection influencing the agent's behavior is negligible.
Persistence and Privilege Escalation (SAFE): No mechanisms for persistence (cron, startup scripts) or privilege escalation (sudo, chmod) are present in the code.

seo-fundamentals