The Agent Skills Directory

[COMMAND_EXECUTION]: The run-pipeline.py script uses subprocess.run to coordinate the execution of multiple sub-scripts and shell commands, such as firecrawl-extract.sh and various build utilities. This behavior is consistent with the skill's primary purpose as an orchestrator.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external websites via a user-supplied URL, creating a surface where malicious instructions on a webpage could influence subsequent LLM-based pipeline stages.
Ingestion points: The --url parameter in run-pipeline.py is used to trigger extraction stages that pull content into the workspace/brand/ directory.
Boundary markers: The orchestrator does not explicitly wrap ingested content in delimiters or provide 'ignore instructions' warnings at this level.
Capability inventory: The skill has the capability to execute shell commands and Python scripts via subprocess.run (in run-pipeline.py) and perform file system operations (in page-admin.py and run-pipeline.py).
Sanitization: The skill uses a slugify function to sanitize file and package names, but it does not sanitize the actual content extracted from the URL before it is used by downstream components.

landing-page-factory-orchestrator