The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes job descriptions from zhipin.com, creating a surface for indirect prompt injection.
Ingestion points: In SKILL.md, the agent is instructed in '主流程 A' to use Ctrl+A and Ctrl+C on the job details pane to retrieve all job description text for evaluation.
Boundary markers: No boundary markers or specific instructions to ignore embedded prompts are provided to the agent to protect against malicious instructions hidden within the job descriptions.
Capability inventory: The agent has capabilities to navigate to arbitrary URLs, click action buttons ('立即沟通', '收藏'), and transmit user-provided messages and resumes.
Sanitization: The skill does not provide any instructions for sanitizing or validating the extracted text content before it is processed by the agent.
[DATA_EXFILTRATION]: The skill automates the transfer of sensitive personal documents to a remote service.
The '发起沟通' (Initiate Communication) workflow involves sending a '微简历' (micro-resume) and an attached resume to the recruitment platform. While this is the primary intended function for a job-seeking agent, the automated handling and transmission of PII (Personally Identifiable Information) is a sensitive operation.

bosszhipin