ljg-plain
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill takes unstructured input from various sources including untrusted URLs via WebFetch and documents via WebSearch/Read, but does not use strict input sanitization or explicit boundaries to prevent prompt injections within the fetched content from altering the agent's behavior.
- [COMMAND_EXECUTION]: The skill uses specific shell-like command formatting for filename timestamps (
date +%Y%m%dT%H%M%S) and output definitions, creating an avenue where complex or altered syntax from raw inputs might interact with internal agent executions if the platform processes these literally as shell expressions.
Audit Metadata