The Agent Skills Directory

Standard Engineering Guidelines: The skill consists of detailed checklists for evaluating correctness, performance, and concurrency. These instructions are designed to improve code quality and do not contain executable logic or malicious patterns.
Security-Conscious Auditing: It specifically instructs the agent to look for safety issues such as improper FFI boundary handling, race conditions, and unsafe Rust patterns. These instructions encourage the development of more secure software.
No Capability Misuse: The skill does not request network access, perform file system operations beyond analysis, or attempt to exfiltrate data. It operates purely as a set of logical constraints and evaluation criteria for the AI agent.
Absence of Obfuscation: The content is entirely transparent, written in plain Markdown, with no hidden commands, encoded strings, or suspicious metadata.

code-review