The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the agent to execute local Node.js management scripts (generate_batch.mjs and export_decisions.mjs) to handle content ingestion and file export tasks. These scripts are provided within the skill package.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided content to generate drafts, which is a core function of its content repurposing capabilities.
Ingestion points: Untrusted content enters the system via the --source parameter (supporting both file paths and raw text) in the scripts/generate_batch.mjs script.
Boundary markers: The instructions explicitly direct the agent to use professional judgment, verify claims, and adhere to a "human-in-the-loop" review workflow. The dashboard acts as a functional boundary between generation and export.
Capability inventory: The skill possesses the capability to write records to a Busabase instance and package processed content into local ZIP archives within the exports/ directory.
Sanitization: The skill implements path traversal protection in scripts/lib/content-assets.mjs by using path.resolve and scope validation to ensure that all asset resolution is confined to the authorized project root.

kelly-writer