The Agent Skills Directory

[PROMPT_INJECTION]: The skill facilitates the processing of untrusted project data (Rust source code) without providing safety boundaries or sanitization instructions for the agent.
Ingestion points: The skill is intended to review, debug, and write Rust code within a project directory (SKILL.md).
Boundary markers: There are no instructions to use delimiters or to ignore potential instructions embedded in code comments or string literals.
Capability inventory: The 'Validation Checklist' and various rule files (e.g., test-tokio-async.md, opt-pgo-profile.md) recommend executing cargo test, cargo clippy, and cargo bench. These commands execute code defined in the project's tests and build scripts.
Sanitization: The skill does not instruct the agent to sanitize or validate project files before processing or executing them.

rust-developer