The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches crate metadata from the official crates.io API and downloads repository source code via git clone. These operations target the official Rust registry and its referenced repositories.
[COMMAND_EXECUTION]: Utilizes a series of shell commands including curl, jq, and git to manage the lifecycle of crate exploration. It creates local directories in /tmp/rust-crates/ and manages repository state via git checkout.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it ingests untrusted metadata (repository URLs and version strings) from an external API and interpolates that data into shell commands executed by the agent.
Ingestion points: Data retrieved from https://crates.io/api/v1/crates/$CRATE_NAME (specifically the .crate.repository and .crate.max_version fields).
Boundary markers: Absent. The skill provides shell templates where external data is inserted directly into command strings.
Capability inventory: File system writes in /tmp, network access via curl and git, and repository manipulation via git checkout.
Sanitization: No explicit validation or sanitization is performed on the fields returned by the API before they are used in shell command construction.

rust-crate-source