The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute shell commands using user-provided input, specifically python scripts/pagespeed_check.py <url> --json, python scripts/crux_history.py <url> --json, and python scripts/gsc_inspect.py <url> --json. This pattern is vulnerable to command injection if the <url> argument is not strictly sanitized or escaped by the agent before shell execution.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external websites (including robots.txt, XML sitemaps, and page content). This creates a surface for indirect prompt injection attacks where a malicious site could influence the agent's behavior or output.
Ingestion points: Data ingested from external URLs, robots.txt, and site metadata in SKILL.md (via scripts and DataForSEO).
Boundary markers: None identified; instructions do not include delimiters or warnings to ignore instructions embedded in the crawled content.
Capability inventory: Shell command execution (Python scripts), file system access (scripts directory), and network access (implied for SEO auditing).
Sanitization: No explicit sanitization or validation steps for the content fetched from external URLs are mentioned.
[DYNAMIC_EXECUTION]: The skill relies on external Python scripts (pagespeed_check.py, crux_history.py, gsc_inspect.py) located in a scripts/ directory that were not included in the provided files. As a result, the internal logic and safety of these executable components cannot be verified.

seo-technical