The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute multiple Python scripts (e.g., validate_seo_findings.py, manage_markdown_site.py, and manage_markdown_work.py) that are not provided within the skill package. While these appear to be part of an intended local toolsuite, executing unverified scripts poses a risk if they are compromised or replaced. It is recommended to verify the integrity of these scripts and apply human review to their outputs.\n- [DYNAMIC_EXECUTION]: Execution paths for the required scripts are resolved at runtime using the CLAUDE_PLUGIN_ROOT environment variable or the .seo-suite-runtime.json configuration file. This dynamic resolution introduces a dependency on the security of the local environment and configuration. Users should ensure that these environment variables and configuration files are stored in protected directories with restricted permissions.\n- [INDIRECT_PROMPT_INJECTION]: The core functionality involves ingesting data from external URLs and page captures, which are inherently untrusted and could contain malicious instructions designed to manipulate the audit findings or subvert the agent's logic. To mitigate this, external content should be wrapped in strict delimiters with explicit 'ignore instructions' warnings.\n
Ingestion points: External URLs, raw/rendered web captures, and SEO findings JSON bundles.\n
Boundary markers: No specific delimiters or isolation markers are provided in the instructions for retrieved external content.\n
Capability inventory: The agent has the capability to execute shell commands (Python scripts) and perform file system operations (writing reports, JSON artifacts, and work indices).\n
Sanitization: There is no mention of sanitization or validation of the ingested external content before it is processed by the LLM or passed to the validation scripts.

seo-page