The Agent Skills Directory

[SAFE]: The skill is designed with privacy-conscious defaults, such as operating within a specific workspace directory (./jinchenma-article-writing-workspace/) and instructing the agent to never read from private directories or write user data to the skill's installation path.
[EXTERNAL_DOWNLOADS]: The skill utilizes network-based tools like Web Search and xcrawl to retrieve research materials. These operations are limited to gathering information for article writing based on user-defined topics.
[COMMAND_EXECUTION]: The skill manages article drafts and research notes through standard file system operations. These actions are confined to the defined project workspace to ensure data organization and prevent unauthorized file access.
[PROMPT_INJECTION]: The skill processes content from the web and user-provided files, representing an indirect prompt injection surface.
Ingestion points: Data enters the agent's context through web search results, crawler output (xcrawl), and user-specified local files mentioned in SKILL.md.
Boundary markers: The skill does not implement technical delimiters for external data but provides clear instructional boundaries.
Capability inventory: The agent possesses file-writing capabilities within the project workspace and network-access capabilities for research.
Sanitization: Instructions require the agent to summarize external content in its own words and avoid direct copying of raw source text, reducing the risk of accidental instruction obedience.

jinchenma-article-writing