fix

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill uses user-controlled input through the $ARGUMENTS variable, including test file paths and test names, which are then interpolated directly into shell commands (e.g., 'npx playwright test '). This creates a vulnerability surface where a malicious user could provide a crafted input containing shell metacharacters to attempt command injection.
  • Ingestion points: The $ARGUMENTS block in SKILL.md, which accepts file paths, test names, and descriptions from the user.
  • Boundary markers: Absent; there are no instructions for the agent to use delimiters or ignore instructions within the user-supplied data.
  • Capability inventory: Subprocess execution via 'npx playwright' commands described in SKILL.md.
  • Sanitization: Absent; the skill does not specify any validation, escaping, or sanitization for the provided arguments before they are used in shell commands.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 02:34 PM
Security Audit — agent-trust-hub — fix