The Agent Skills Directory

[COMMAND_EXECUTION] (HIGH): The skill requires the agent to run cargo test, cargo clippy, and cargo fmt. These commands execute code within the project's context, allowing for arbitrary code execution on the host machine.
[PROMPT_INJECTION] (HIGH): (Category 8
Indirect Prompt Injection): The agent is vulnerable to malicious instructions embedded in the code it processes. Evidence: 1. Ingestion points: Rust source files and command outputs from tools like git, docker, and kubectl. 2. Boundary markers: Absent. 3. Capability inventory: Code execution via cargo commands. 4. Sanitization: Absent.
[REMOTE_CODE_EXECUTION] (HIGH): The ability to execute tests and build code derived from untrusted project files provides an attack vector for code execution if the repository being worked on is malicious.
[DATA_EXPOSURE] (LOW): The patterns include checking environment variables and system paths (e.g., /etc/passwd). While intended for defensive testing, the skill's logic confirms the agent can access and process sensitive strings from the environment.

rtk-tdd