The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill determines the path for its utility scripts at runtime by reading the suite_root value from a local .seo-suite-runtime.json file or by using the CLAUDE_PLUGIN_ROOT environment variable. This path is then used to construct shell commands for execution.
Evidence: Instructions in SKILL.md and references/markdown-delivery-contract.md describe how to resolve <suite-root> and use it to execute scripts located in the scripts/ directory.
[COMMAND_EXECUTION]: The skill executes multiple shell commands to run Python scripts for task validation, site management, and workflow initialization.
Evidence: Commands such as python "<suite-root>/scripts/validate_seo_findings.py", python "<suite-root>/scripts/manage_markdown_site.py", and python "<suite-root>/scripts/manage_markdown_work.py" are called across SKILL.md and references/markdown-delivery-contract.md.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, such as rendered web pages and asset inventories, which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: The skill requires rendered pages, asset inventories, and source URLs as inputs (SKILL.md).
Boundary markers: There are no instructions defining specific delimiters or warnings to ignore instructions embedded within the ingested external data.
Capability inventory: The skill has the capability to write various project files (e.g., image-review.md, index.md, seo-findings.json) and execute local Python scripts (SKILL.md, references/markdown-delivery-contract.md).
Sanitization: The instructions do not specify any methods for sanitizing, escaping, or filtering the external content before it is processed or used in script arguments.

seo-images