The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to execute code downloaded at runtime using the npx package runner. Commands like npx -y @nomadamas/k-skill@0 instruct download and execute scripts from the npm registry, which is a form of remote code execution.
[EXTERNAL_DOWNLOADS]: Core instructions and auxiliary files are fetched from external sources, specifically the npm registry and a GitHub repository (github.com/NomaDamas/k-skill), rather than being entirely self-contained.
[COMMAND_EXECUTION]: The skill utilizes shell commands (npx) to perform updates, fetch instructions, and read documents.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze candidate profiles and resumes from saramin.co.kr. This creates a vulnerability where instructions embedded in a resume could influence the agent's behavior.
Ingestion points: Candidate profiles and resumes accessed via browser automation on saramin.co.kr as defined in instruction.md.
Boundary markers: The instructions do not specify the use of delimiters or warnings to ignore embedded instructions in the ingested data.
Capability inventory: The agent can navigate the web, read candidate information, and generate recruitment shortlists.
Sanitization: There is no evidence of sanitization or filtering applied to the external resume content before processing.

saramin-talent-search