The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The SKILL.md file instructs the agent to run npx -y @nomadamas/k-skill@0 instruct hankookilbo-news. This pattern downloads and executes code from the NPM registry at runtime to provide dynamic instructions.
[COMMAND_EXECUTION]: The instruction.md file contains examples and workflows that require the execution of shell commands, specifically curl for network requests and python3 -c for processing JSON data via a one-liner script.
[EXTERNAL_DOWNLOADS]: The skill connects to an external endpoint at https://mcp.hankookilbo.com/mcp to fetch news metadata and horoscope information. This communication happens via curl without additional authentication.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks (Category 8). It ingests untrusted data from the Hankook Ilbo MCP server (news titles and excerpts) which are then processed by the agent.
Ingestion points: News metadata items from https://mcp.hankookilbo.com/mcp (instruction.md).
Boundary markers: The instructions include logical boundaries such as '제목은 item.title 원문을 그대로 인용한다' (quote titles verbatim) and '발췌에 없는 해석... 단정하지 않는다' (do not conclude based on interpretations not in the excerpt).
Capability inventory: Uses curl for network access and python3 for script execution (instruction.md).
Sanitization: There is no explicit sanitization or filtering of the external content before it is processed or displayed.

hankookilbo-news