AGENT LAB: SKILLS

fal-workflow

Fail

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [REMOTE_CODE_EXECUTION] (HIGH): The script scripts/create-workflow.sh is vulnerable to a Python injection attack. In the script, the shell variable $NODES is directly interpolated into a triple-quoted Python string (nodes = json.loads('''$NODES''')). An attacker can provide a payload containing triple quotes (''') to escape the string literal and execute arbitrary Python commands on the host system.
  • [COMMAND_EXECUTION] (HIGH): The skill instructs the AI agent to execute a local bash script (/mnt/skills/user/fal-workflow/scripts/create-workflow.sh) using arguments derived from user input. Because the script lacks input sanitization or safe parsing mechanisms for these arguments, it creates a direct path for the execution of malicious payloads.
  • [COMMAND_EXECUTION] (MEDIUM): The script scripts/create-workflow.sh also demonstrates unsafe JSON construction. It uses a bash heredoc to interpolate shell variables ($NAME, $TITLE, $DESCRIPTION) directly into a JSON template. This allows an attacker to inject additional JSON fields or corrupt the structure of the generated workflow file.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 17, 2026, 04:31 PM