dingtalk-hrbrain

Pass

Audited by Gen Agent Trust Hub on Jul 24, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill interacts with the dws binary to perform HR-related tasks, such as querying talent pools and employee archives via the hrbrain command prefix.
  • [PROMPT_INJECTION]: The skill processes untrusted user input (e.g., search keywords and JSON query objects) which are passed as arguments to CLI commands, creating a surface for indirect prompt injection.
  • Ingestion points: Untrusted data enters the context through parameters like --keyword, --origin-json, and --data-queries as defined in SKILL.md.
  • Boundary markers: None identified in the instruction set to separate command logic from data input.
  • Capability inventory: The skill utilizes dws hrbrain subcommands to read, query, and search sensitive organizational personnel data.
  • Sanitization: While the skill specifies that certain parameters must be valid JSON strings, it does not explicitly define sanitization or escaping of the values within those structures.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 24, 2026, 07:07 PM
Security Audit — agent-trust-hub — dingtalk-hrbrain