The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes build-related shell commands such as cargo build, npm run build, and python setup.py. These are initiated within the BuildManager class in scripts/run_build-manager.py to facilitate project compilation and deployment.
[EXTERNAL_DOWNLOADS]: The skill interacts with external registries including crates.io, registry.npmjs.org, and github.com to resolve and fetch project dependencies. These are well-known, official services for the supported languages (Rust, Node.js) and the download findings are considered safe under their intended use.
[DYNAMIC_EXECUTION]: The Python script scripts/run_build-manager.py dynamically constructs and executes shell commands using subprocess.Popen and subprocess.run. This is used to manage build progress, capture output, and implement hot-reloading functionality for compiled binaries.
[INDIRECT_PROMPT_INJECTION]: The skill processes project configuration files (e.g., package.json, Cargo.toml) and source code to determine build logic. This ingestion of untrusted data into an environment with command execution capabilities represents a potential attack surface, although it is constrained by the build tool context.
Ingestion points: Reads Cargo.toml, package.json, and source files (.rs, .js, .py) via codex_read_file and globbing in scripts/run_build-manager.py.
Boundary markers: None identified in the prompt interpolation or tool usage.
Capability inventory: High-privilege command execution via subprocess calls for cargo, npm, and direct binary execution in perform_hot_reload.
Sanitization: No specific sanitization or validation of build configuration field values before inclusion in shell commands.

build-manager