The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. 1. Ingestion points: The agent reads external web content via web search and the xcrawl scraping tool, as well as local files from user-specified paths. 2. Boundary markers: No explicit delimiters are used for ingested content to prevent embedded instructions from affecting the agent context. 3. Capability inventory: The skill has access to file system operations (read/write), web search, and web scraping tools. 4. Sanitization: The skill instructs the agent to summarize content in its own words and avoid verbatim copying, but no technical filtering or escaping is applied to the ingested data.
[EXTERNAL_DOWNLOADS]: The skill performs automated web retrieval using search engines and the xcrawl scraping tool to download and process content from external URLs provided by the user or discovered via search.
[DATA_EXFILTRATION]: The skill accesses local files and directories based on user input. While intended for processing writing materials, this capability allows the agent to read any file the user provides a path for, which could include sensitive information if the agent is manipulated or the user provides a sensitive path.

jcm-article-writing