The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The Cargo.lock file in the examples/golden-performance/ directory contains suspicious dependency entries that do not align with official Rust ecosystem standards. Specifically, the well-known serde_json package (version 1.0.151) is listed with an unusual dependency on a package named zmij, and serde (version 1.0.229) is shown depending on serde_core. These versions and dependency structures are non-existent or impersonate legitimate libraries, indicating a supply chain attack vector.
[REMOTE_CODE_EXECUTION]: The skill project is configured to download and execute untrusted code if the included example is built or tested. Using tools like cargo build or cargo test as suggested in the skill's workflow would trigger the resolution of the poisoned Cargo.lock file, leading to the execution of the suspicious zmij and serde_core crates within the user's environment.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) as it instructs the agent to ingest and analyze untrusted external data, such as Rust source code and performance profile results.
Ingestion points: User-provided source code, benchmark outputs, and performance profiles (e.g., heap dumps, flame graphs) in SKILL.md.
Boundary markers: The instructions lack explicit delimiters or warnings to ignore instructions embedded within the analyzed data.
Capability inventory: The skill encourages the use of powerful CLI tools (cargo, perf, samply) and benchmark script execution.
Sanitization: There is no evidence of input validation or sanitization for the data processed during profiling and optimization steps.

rust-performance