The Agent Skills Directory

[PROMPT_INJECTION]: The skill audits external websites by fetching and parsing HTML content (e.g., scripts/seo_audit.py), which introduces a surface for indirect prompt injection. Malicious instructions embedded in a target website's meta tags or schema could attempt to influence the agent's analysis.
Ingestion points: Website content fetched via scripts/seo_audit.py and curl commands.
Boundary markers: The skill does not define specific delimiters to separate untrusted web content from its own instructions.
Capability inventory: The toolkit includes network access and keyword analysis scripts.
Sanitization: No filtering or sanitization is applied to the fetched HTML content before it is processed by the agent.
[DATA_EXFILTRATION]: Several scripts (such as scripts/keyword_research.py) transmit query data (keywords, domains) to api.dataforseo.com. This is a standard integration with a well-known SEO data service and aligns with the skill's functional requirements.
[REMOTE_CODE_EXECUTION]: The documentation in references/schema-templates.md contains a curl | bash command pattern inside a JSON-LD FAQ example. This is used solely as a technical placeholder for documentation and does not instruct the agent to execute code from a remote source.

hebrew-seo-geo-toolkit