The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to analyze external Rust codebases and Tauri configuration files, which inherently introduces a surface for indirect prompt injection. Maliciously crafted code snippets provided for review could contain instructions aimed at subverting the agent's behavior. However, the risk is categorized as safe because the skill has no executable capabilities (such as shell access, network operations, or filesystem writes) that could be exploited by such an injection.
Ingestion points: The skill analyzes src-tauri/**/*.rs files, tauri.conf.json, Cargo.toml, and user-pasted code fragments.
Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded commands within the analyzed code, but this is consistent with a knowledge-based review tool.
Capability inventory: The skill is restricted to a knowledge-based model with no network access, no shell execution, and no filesystem write permissions as defined in permissions.yaml.
Sanitization: No specific sanitization of the audited code content is performed, but the lack of active tools prevents exploitation.
[REMOTE_CODE_EXECUTION]: The skill documentation (specifically references/gates.md) recommends the use of standard Rust security and quality tools such as cargo audit and cargo machete. These are industry-standard tools for dependency auditing and are provided as recommendations for the user to implement in their own development environment, rather than being executed by the skill itself.
[SAFE]: The skill references numerous external resources for documentation purposes, including the official Rust Book (doc.rust-lang.org), Tauri documentation (v2.tauri.app), and reputable GitHub repositories for Rust tooling (tauri-specta, cargo-machete). All external links are to well-known, trusted documentation and open-source project sites.

tauri-stinky