skills/lijigang/ljg-skills/ljg-learn/Gen Agent Trust Hub

ljg-learn

Pass

Audited by Gen Agent Trust Hub on Oct 26, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied input (the concept name) and interpolates it directly into a file path for local storage. This creates a potential attack surface where a malicious concept name containing shell metacharacters could attempt to influence file operations or command execution.
  • Ingestion points: The concept name provided by the user (e.g., via the /ljg-learn trigger) is used in the output filename.
  • Boundary markers: None. The user input is directly embedded into the instruction for file creation.
  • Capability inventory: The skill requires the ability to execute shell commands (date) and write files to the filesystem (~/Documents/notes/).
  • Sanitization: The instructions do not specify any validation or sanitization for the user-provided concept name before it is used in the filename template.
  • [COMMAND_EXECUTION]: The skill explicitly instructs the agent to run the shell command date +%Y%m%dT%H%M%S to generate a timestamp. While the command itself is benign, it establishes a pattern of shell usage that could be targeted if combined with unsanitized user input.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 26, 2026, 01:23 PM