The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the user to download and execute an installation script directly from the official Wasmtime project domain.
Evidence: curl https://wasmtime.dev/install.sh -sSf | bash in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for ingesting and processing external WebAssembly (.wasm) modules, which represents a potential attack surface for indirect prompt injection or logic manipulation if modules are sourced from untrusted parties.
Ingestion points: Loading WASM modules from the filesystem via the CLI (wasmtime prog.wasm) or via Rust code (Module::from_file(&engine, "prog.wasm")).
Boundary markers: None explicitly specified for incoming module data.
Capability inventory: The skill utilizes file system access (via pre-opened directories), environment variable inheritance, and arbitrary code execution within the WASM sandbox.
Sanitization: The skill provides proactive security guidance, demonstrating how to use fuel metering (config.consume_fuel(true)) to limit CPU execution and directory mapping (--dir /tmp::/) to enforce filesystem sandboxing.
[DYNAMIC_EXECUTION]: The core functionality of the skill involves loading, compiling, and executing binary WebAssembly modules at runtime.
Evidence: Documentation of wasmtime run, Module::from_file, and Instance::new for executing compiled code.
[COMMAND_EXECUTION]: The skill recommends installing development toolchain components using standard package managers.
Evidence: cargo install wasm-tools cargo-component in SKILL.md.

wasm-wasmtime