The Agent Skills Directory

[PROMPT_INJECTION]: The ai_screening logic in SKILL.md contains a potential surface for indirect prompt injection where untrusted external data is processed by the model.
Ingestion points: The {resume_text} variable in the screening prompt interpolates content from resumes provided by job applicants directly into the model's instruction context.
Boundary markers: The prompt template lacks explicit delimiters or clear instructions for the model to ignore any embedded commands or formatting within the resume text.
Capability inventory: The skill is connected to HR tools (greenhouse_api, workday_api, bamboohr_api) via the hr-mcp server and has the capability to perform automated hiring actions such as auto_reject or advancing candidates to subsequent stages.
Sanitization: There is no evidence of sanitization, filtering, or validation of the input variables before they are used in the automated screening workflow.

hr-automation