The Agent Skills Directory

[COMMAND_EXECUTION]: The skill references standard Rust development commands, specifically cargo test, cargo miri, and cargo clippy. These are legitimate tools for testing and linting code in a development environment and are appropriate for the skill's stated purpose of verifying Rust code safety.
[EXTERNAL_DOWNLOADS]: The instructions suggest using established third-party crates like bytemuck, zerocopy, arrayvec, and smallvec. These are widely recognized, community-vetted, and trusted libraries within the Rust ecosystem for handling memory-related tasks safely, and the skill correctly identifies them as safer alternatives to writing custom unsafe code.
[PROMPT_INJECTION]: The content does not contain instructions that attempt to bypass AI safety guardrails, override system-level behavior, or extract system prompts. The language is purely instructional and focused on Rust programming best practices.
[DATA_EXFILTRATION]: There are no patterns indicating the collection or transmission of sensitive local data, environment variables, or credentials. The network surface is minimal and restricted to standard package management or tool installation expectations.
[INDIRECT_PROMPT_INJECTION]: The skill provides a framework for analyzing user-provided code (untrusted data). It encourages the agent to look for specific safety comments (// SAFETY:) and documentation sections, which acts as a structured review process. The capability tier is limited to standard local development tools.

unsafe-rust