The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions in SKILL.md direct the agent to execute curl commands using variables like <视频直链> which are parsed from external web pages. This creates a risk of command injection if the scraped content contains shell metacharacters and the agent does not properly escape the input before execution.
[PROMPT_INJECTION]: Indirect prompt injection surface identified. The skill ingests untrusted data from Xiaohongshu (notes, ASR transcripts, OCR text) and processes it with an LLM to generate summaries and structure analysis for database entry.
Ingestion points: scripts/fetch.py fetches data and metadata from Xiaohongshu note URLs.
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are present in the instructions to protect the LLM from malicious content within the scraped notes.
Capability inventory: Subprocess execution (python3, curl), Network access (feishu.cn API), and File system writing.
Sanitization: No explicit sanitization, escaping, or validation of the external content is described before it is passed to the LLM or used in shell commands.
[EXTERNAL_DOWNLOADS]: The scripts/fetch.py script requires the requests library. While a common and well-known package, it is a dependency that must be installed from an external registry.
[DATA_EXFILTRATION]: The skill accesses sensitive local configuration paths at $HOME/.config/agent-secrets/ to retrieve API tokens and environment variables. While intended for legitimate authentication to Feishu (a well-known service), this pattern involves the automated retrieval and transmission of sensitive credentials.

xiaohongshu-fetch