The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of arbitrary shell commands through the --parallel-tests argument in the scripts/autoreview script. These commands are executed via subprocess.Popen with shell=True, which could lead to arbitrary code execution if the input command is maliciously controlled.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection. It includes unvalidated repository content, such as git diffs and untrusted files, directly into the prompts sent to external AI review engines. An attacker could craft code that contains malicious instructions designed to deceive the reviewer or suppress findings.
[DATA_EXFILTRATION]: The skill reads repository contents, including source code, diffs, and untracked files, and transmits them to external AI service providers (e.g., OpenAI, Anthropic) for analysis. This is a primary function of the skill but involves the transfer of potentially sensitive data to third-party services.
[COMMAND_EXECUTION]: The scripts/test-review-harness.ps1 script utilizes the -ExecutionPolicy Bypass flag when invoking PowerShell. This bypasses local security policies that might otherwise restrict the execution of PowerShell scripts.

jb-autoreview