The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) because it is designed to ingest and act upon untrusted external data (the code being reviewed).
Ingestion points: The skill uses Read, Glob, and Grep to ingest the contents of files within the target directory as specified in the argument-hint and workflow steps.
Boundary markers: The instructions lack requirements for the agent to use XML delimiters, triple backticks, or specific "ignore embedded instructions" warnings when processing the code files.
Capability inventory: The skill has significant capabilities including Write, Edit, AskUserQuestion, and integration with the GitHub MCP for posting pull request comments, which could be abused if malicious instructions are embedded in the reviewed code.
Sanitization: No sanitization or validation logic is defined to filter natural language instructions out of the code files being analyzed.
[EXTERNAL_DOWNLOADS]: The skill references several third-party GitHub repositories for MCP servers that are not from established or well-known technology organizations.
Evidence: The 'Connected MCP Servers' section lists github.com/Coding-Solo/godot-mcp, github.com/CoplayDev/unity-mcp, and github.com/chongdashu/unreal-mcp as tools for live code analysis.
Risk: While these tools are relevant to the skill's purpose, they are community-maintained resources that require manual verification by the user before installation.

game-code-review