The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill resolves its core execution path (<suite-root>) dynamically from the environment or a local configuration file, which is then used to locate and run Python scripts. This dynamic path resolution allows for the execution of code from computed locations.
Evidence: The instructions in SKILL.md and references/markdown-delivery-contract.md specify reading the .seo-suite-runtime.json file or using the ${CLAUDE_PLUGIN_ROOT} environment variable to determine the path for calling scripts like validate_seo_findings.py and manage_markdown_site.py.
[COMMAND_EXECUTION]: The skill frequently invokes shell commands to execute Python scripts for data processing and validation.
Evidence: Several commands are defined, such as python "<suite-root>/scripts/manage_markdown_site.py" build --inventory <private/inventory.json> --output <new-public-candidate> and python "<suite-root>/scripts/manage_markdown_work.py".
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data sources, which could contain adversarial content meant to influence the agent's logic or output.
Ingestion points: Sitemaps, index files, crawler exports, and CMS lists are listed as primary inputs in the "Input gate" and "Workflow" sections of SKILL.md.
Capability inventory: The skill has the ability to execute multiple Python scripts and perform file operations.
Boundary markers: There are no instructions provided to wrap these external inputs in delimiters or use specific markers to ignore embedded instructions.
Sanitization: The instructions do not specify any validation or sanitization steps for the external data before it is processed by the workflow steps.

seo-sitemap