ios-code-review

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill performs analysis on untrusted data (Swift source code), which introduces a surface for indirect prompt injection. Attackers could embed malicious instructions in code comments or metadata that attempt to override the reviewer's judgment or the resulting review report.\n
  • Ingestion points: User-provided file changes accessed via git diff (SKILL.md).\n
  • Boundary markers: Absent; there are no specific delimiters defined to isolate user code from the agent's instructions.\n
  • Capability inventory: Static analysis tools including git and grep (SKILL.md).\n
  • Sanitization: Absent; the skill does not specify any pre-processing or sanitization of the code before review.\n- [COMMAND_EXECUTION]: The skill provides automated shell commands (git, grep) to facilitate code inspection. While these are standard development tools used for read-only static analysis, they represent a capability that interacts with the underlying operating system.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 07:29 AM