The Agent Skills Directory

[COMMAND_EXECUTION]: The skill frequently executes system commands using the subprocess module. For instance, aggregator/fetch_all.py uses curl to fetch trending data from various platforms. topic_scorer.py executes the pass password manager utility to retrieve API keys for DeepSeek and GLM. Furthermore, draft_reviewer.py and topic_presenter.py execute an external script, newsbot_send.py, to send notifications and cards to the user.
[CREDENTIALS_UNSAFE]: To perform automated publishing and fetching, the skill reads sensitive local files. Scripts like auto_publisher.py and fetch_all.py access browser cookies and session state files stored in ~/.playwright-data/ or ~/.xiaohongshu/ for services like Xiaohongshu and WeChat. It also attempts to retrieve AI provider credentials through the system's pass tool.
[PROMPT_INJECTION]: The skill processes untrusted external data, creating an indirect prompt injection surface.
Ingestion points: aggregator/fetch_all.py scrapes hot topic titles, summaries, and descriptions from platforms like GitHub, Reddit, and Weibo.
Boundary markers: Absent. content_generator.py and topic_scorer.py interpolate fetched strings directly into LLM prompts using simple string replacement.
Capability inventory: The skill has the capability to write files (drafts), perform network requests (to LLM providers and social platforms), and publish content to Xiaohongshu via auto_publisher.py.
Sanitization: No sanitization or filtering of fetched content is performed before it is processed by the AI models.
[EXTERNAL_DOWNLOADS]: The aggregator/fetch_all.py script performs numerous HTTP GET requests to retrieve information from a wide array of sources, including GitHub Trending, YouTube feeds, Bilibili APIs, and several regional news aggregators.

content-factory