The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions advise the agent to execute Python code via shell heredocs (python3 << 'PY') to interact with the provided Ghost API library.
[EXTERNAL_DOWNLOADS]: The ghost_api.py module fetches image metadata and attribution details from Unsplash (unsplash.com). Unsplash is a well-known service.
[EXTERNAL_DOWNLOADS]: The ghost_md.py script utilizes the uv tool to dynamically manage and install well-known Python packages (markdownify, markdown) for content conversion.
[DATA_EXFILTRATION]: The skill provides an upload method in the Ghost class that reads local files and transmits them to the user-configured Ghost API endpoint. This is a core functionality for managing blog media.
[PROMPT_INJECTION]: The skill processes untrusted data from the blog (such as post content and member info) which presents an indirect prompt injection surface.
Ingestion points: Blog post content and metadata retrieved via ghost_api.py and ghost_md.py.
Boundary markers: No explicit delimiters are suggested in the provided code examples to separate external content from instructions.
Capability inventory: File read/write access, network operations to Unsplash and the Ghost API, and destructive actions like post deletion.
Sanitization: Content is converted between formats (HTML/Markdown/Lexical) but no specific instruction filtering is performed.

ghost-blog