The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The script scripts/export_chatgpt.py downloads HTML content from chatgpt.com share URLs using the urllib.request library. Since these downloads target a well-known service (OpenAI), the network operation is considered a neutral function of the skill.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection vulnerability surface because it imports untrusted chat history from external URLs into the agent's workspace.
Ingestion points: Content is ingested via the fetch_html function in scripts/export_chatgpt.py from user-provided URLs.
Boundary markers: The script employs dynamic Markdown code fences (via compute_fence) to encapsulate and isolate message content, which helps distinguish it from other instructions.
Capability inventory: The script has the capability to perform network GET requests and write data to the local file system.
Sanitization: No validation or sanitization of the conversation text is performed; the skill performs a direct export of the remote content.

export-chatgpt