The Agent Skills Directory

[SAFE]: The skill consists of instructions directing the agent to perform a code review on Convex framework code. It does not define any tools, execute shell commands, perform network requests, or access sensitive files. The instructions focus on manual verification of authentication, performance indexing, and code quality patterns.
[INDIRECT_PROMPT_INJECTION]: While the skill is designed to process external code (which could contain malicious instructions), it lacks the capabilities (such as network access, file writing, or command execution) required for an injection attack to perform harmful actions beyond providing a misleading review. This is considered a safe use case for code analysis.

convex-reviewer