The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command using a local script path: "${CLAUDE_PLUGIN_ROOT}/scripts/claude-seo" run metadata_template.py --title "<title>" --description "<desc>" --json.
The arguments <title> and <desc> are populated with content fetched directly from the external URL being analyzed.
This presents a risk of command injection if the page content contains characters used to break out of shell quoting (e.g., backticks, semicolons, or dollar-sign expressions) and if the underlying execution environment does not provide strict argument sanitization.
[INDIRECT_PROMPT_INJECTION]: The skill's primary purpose is to ingest and analyze untrusted data from the open web (supplied URLs).
Ingestion points: Page content, title tags, and meta descriptions are read from external URLs into the agent's context (SKILL.md).
Boundary markers: The instructions do not define clear delimiters or "ignore instructions" warnings when processing the fetched HTML content.
Capability inventory: The agent has the capability to execute shell commands via the claude-seo script and potentially use DataForSEO MCP tools (SKILL.md).
Sanitization: There is no mention of sanitizing or escaping the fetched content before it is processed or passed to command-line arguments.

seo-page