The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill uses cargo generate with the --allow-commands flag to fetch a project template from the author's repository (https://github.com/xsyetopz/rust-template). This allows the execution of a post-generation rhai script on the user's machine. The skill instructions mitigate this risk by requiring the agent to display the hook script (tools/cargo-generate.rhai) to the user for review before passing the execution flag.- [COMMAND_EXECUTION]: The workflow involves executing local Python scripts (tools/bootstrap.py and tools/check.py) to handle project setup and policy enforcement. These scripts are part of the template structure and are run directly by the agent.- [EXTERNAL_DOWNLOADS]: The skill downloads project templates from the author's GitHub repository and installs the cargo-deny tool using cargo install from official registries.- [INDIRECT_PROMPT_INJECTION]: The skill's logic flows based on the content of files in the current working directory, which could be influenced by an attacker to manipulate agent behavior.
Ingestion points: The agent is instructed to read Cargo.toml for workspace.metadata.rust-policy keys and check for the existence of tools/bootstrap.py to determine its context (SKILL.md).
Boundary markers: Absent; the agent relies on the presence or absence of specific file structures without explicit delimiters for untrusted content.
Capability inventory: Extensive capabilities including shell command execution (python3, cargo, bun, rustup) and file system writes.
Sanitization: Absent; configuration data from the project files directly influences the command-line arguments used in generation and extension tasks.

scaffold-strict-rust-projects