commit-review

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION] (HIGH): The skill is vulnerable to Indirect Prompt Injection because it ingests untrusted code data to perform its primary function.\n
  • Ingestion points: The skill instructs the agent to read external code changes using git diff (found in the '评审流程' section of SKILL.md).\n
  • Boundary markers: Absent. There are no instructions or delimiters defined to separate the instructions of the skill from the content of the code being reviewed.\n
  • Capability inventory: The agent executes shell commands (git) and makes critical security/quality decisions that influence the development pipeline.\n
  • Sanitization: Absent. The skill lacks any mechanism to sanitize or escape the content of the code diffs before processing them.\n- [COMMAND_EXECUTION] (MEDIUM): The skill requires the execution of system commands (git diff, git diff --cached) to operate. If an agent is directed to a malicious repository, these commands or the subsequent processing of their output could be exploited to compromise the agent's logic.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 12:43 PM