review-changes
Pass
Audited by Gen Agent Trust Hub on May 20, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill interpolates user-supplied input (
$ARGUMENTS) directly into the prompt template for the 'reviewer' subagent. The instructionprompt: "Review changes. Arguments: $ARGUMENTS"lacks boundary markers or escaping, allowing a user to potentially override the subagent's behavior through crafted arguments. - [PROMPT_INJECTION]: The skill is designed to process untrusted data in the form of branch changes and pull requests. This represents an indirect prompt injection surface where an attacker could embed malicious instructions within code comments or documentation to influence the agent's output during the review process.
- Ingestion points: User arguments (
$ARGUMENTS) and external code changes (implicitly via the 'reviewer' subagent functionality). - Boundary markers: Absent. The template does not use delimiters (like XML tags or triple quotes) to isolate untrusted data from the system prompt.
- Capability inventory: Dispatches execution to a subagent ('reviewer').
- Sanitization: None detected. The skill passes input directly into the prompt string without filtering or validation.
Audit Metadata