The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed around the execution of multiple local Python scripts (e.g., boss_login.py, boss_greet.py, chat_engine.py, resume_parser.py) to automate browser interactions and data processing.
[DATA_EXFILTRATION]: The skill includes functionality to upload extracted resume information to an external platform (Feishu/Lark) via scripts/feishu_upload.py using configuration from config/feishu.json.
[CREDENTIALS_UNSAFE]: The documentation explicitly mentions the storage of sensitive credentials in local files, including:
config/llm.json: Contains api_key for LLM providers.
config/feishu.json: Contains app_id and table_id for Feishu integration.
data/cookies/boss_cookies.json: Stores authentication cookies for Boss Zhipin, which could be used for session hijacking if compromised.
[PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection as it processes untrusted data from external sources:
Ingestion points: scripts/resume_parser.py ingests content from PDF/DOCX resumes, and scripts/chat_engine.py ingests incoming chat messages from candidates.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are mentioned in the documentation.
Capability inventory: The agent has the capability to execute shell commands, read/write local files, and make network requests to LLM and Feishu APIs.
Sanitization: There is no evidence of sanitization or filtering of the content extracted from resumes or chat messages before it is passed to the LLM for decision-making.

openhr