The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it fetches and processes untrusted raw HTML from external URLs.
Ingestion points: scripts/fetch_page.py and scripts/parse_html.py retrieve and parse content from user-provided product URLs.
Boundary markers: None identified; the skill instructions do not specify delimiters or instructions to ignore embedded content.
Capability inventory: Execution of local scripts for data normalization and API interaction.
Sanitization: No sanitization or filtering of external HTML is described in the workflow.
[COMMAND_EXECUTION]: The skill makes extensive use of local script execution to perform its primary SEO and marketplace analysis tasks.
[EXTERNAL_DOWNLOADS]: The skill fetches content from arbitrary external websites for on-page SEO auditing and communicates with the DataForSEO Merchant API.
[DATA_EXFILTRATION]: The ucp_check.py script includes a --probe-endpoints feature that performs HEAD requests to URLs defined in a remote /.well-known/ucp JSON file. This mechanism could be leveraged for Server-Side Request Forgery (SSRF) if a malicious profile directs the agent to probe internal or non-public network resources.

seo-ecommerce