The Agent Skills Directory

Indirect Prompt Injection (HIGH): The skill is designed to ingest and analyze untrusted external Rust/Wasm source code.
Ingestion points: Processes external application code and dependency trees.
Boundary markers: Absent in the instruction set.
Capability inventory: Recommends running cargo clippy, cargo geiger, and cargo fuzz on target projects. These tools trigger the execution of project-defined build.rs scripts and procedural macros during compilation or analysis.
Sanitization: None specified.
Risk: An attacker-controlled project can achieve Remote Code Execution (RCE) by including malicious logic in build.rs, which the agent would execute when following the skill's auditing instructions.

security-audit