The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests resumes from an external directory provided via the --batch argument, which creates an attack surface for indirect prompt injection. Malicious instructions embedded within a resume could attempt to influence the agent's behavior. However, the skill implements a comprehensive multi-step validation process, including a semantic-intent validator (Step 6) and a mandatory human-in-the-loop review (Step 9), which provides high-confidence mitigation against such risks.
[DATA_EXPOSURE]: The skill handles high volumes of sensitive PII (resumes) as its primary purpose. It minimizes exposure risk by implementing a dedicated proxy scanner that redacts or normalizes 36 categories of identity markers before the data is processed by the language model.
[SAFE]: Analysis of the instructions and metadata confirms that the skill contains no executable code, no remote script dependencies, and no network operations. The system functions as a descriptive extraction tool with strict non-evaluative constraints.

resume-summarizer