skills/allenai/asta-plugins/do-work/Gen Agent Trust Hub

do-work

Pass

Audited by Gen Agent Trust Hub on Jul 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill is configured to run commands found within the '# Instructions' section of a 'README.md' file using the Bash tool. This provides a direct path from file content to shell execution.
  • [PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by executing instructions from 'work//README.md' as a primary procedure.
  • Ingestion points: The '# Instructions' section of 'work//README.md' is the primary source of external instructions.
  • Boundary markers: There are no identified delimiters or isolation techniques used to prevent the instructions from overriding agent behavior.
  • Capability inventory: The skill has access to powerful capabilities including 'Bash(*)', file modifications via 'Write(work/**)', and several 'asta-tools' and 'asta-assistant' skills.
  • Sanitization: No input validation or sanitization is performed on the instructions before they are executed.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 7, 2026, 05:19 PM
Security Audit — agent-trust-hub — do-work