The Agent Skills Directory

[COMMAND_EXECUTION]: The skill triggers the execution of local Python utilities (blog_render.py and blog_preflight.py) to perform content rendering and pre-publication audits. It enforces security best practices by requiring that the script directory (BLOG_SCRIPT_DIR) be resolved to an absolute path and refusing execution if non-absolute overrides are provided.
[DYNAMIC_EXECUTION]: The skill assembles shell command strings at runtime to invoke the rendering and preflight scripts, incorporating dynamic variables such as <localized-slug> into the command arguments.
[INDIRECT_PROMPT_INJECTION]: The skill has a documented indirect prompt injection surface due to its nature as a multi-agent orchestrator.
Ingestion points: The skill ingests blog content (from SKILL.md, Phase 2) and passes it through translation (Phase 3), localization (Phase 4), and review (Phase 6) agents.
Boundary markers: There are no explicit delimiters or specific 'ignore embedded instructions' warnings provided in the orchestration logic to prevent the agent from obeying instructions hidden within the blog content.
Capability inventory: The workflow includes writing files to the multilingual/ directory and executing subprocesses via python3 for rendering and validation.
Sanitization: No explicit sanitization or escaping of the processed blog text is described before it is passed to downstream agents or scripts.

blog-multilingual