The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill processes untrusted external data from Phase 2 Implementation Plans and Phase 2.5 Interview Findings. It uses this content to generate executable code and tests without specified boundary markers or sanitization.
[Command Execution] (HIGH): The agent is instructed to run cargo test, cargo audit, and cargo tarpaulin. These commands execute the code generated by the agent on the local system. An attacker can inject malicious logic into the implementation plan that executes during the test or audit phase.
[Dynamic Execution] (MEDIUM): The workflow involves runtime generation, compilation, and execution of Rust code based on instructions derived from external documents, which is a significant attack surface.

disciplined-implementation