The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to fetch the 'uv' toolchain installer from Astral's official domain (astral.sh) to initialize the development environment.
[REMOTE_CODE_EXECUTION]: Installation scripts in 'scripts/install.md' and 'references/local-toolchain/setup-native.md' direct the execution of a remote shell script via 'curl | sh' or 'wget | sh' for environment bootstrapping.
[PRIVILEGE_ESCALATION]: Documentation across several files (e.g., 'scripts/debug.md', 'scripts/dev-setup.md') instructs the agent or user to use 'sudo' for installing system-level packages such as cross-compilers and for accessing serial communication devices for firmware logging.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to manage and interpret ZMK project files (.keymap and .conf). This creates a vulnerability surface where malicious instructions could be embedded in code comments within these files, which an agent might follow while performing analysis or build tasks.
Ingestion points: Processes external user-supplied data in '.keymap' and '.conf' configuration files (referenced in 'references/config/overview.md').
Boundary markers: Absent. There are no instructions for the agent to distinguish between configuration data and potentially malicious natural language instructions embedded in comments.
Capability inventory: Extensive shell command execution (via 'west', 'git', 'uv', and 'zmk' CLI tools), file writing capabilities, and system-level operations via 'sudo'.
Sanitization: None. The skill does not provide mechanisms to sanitize or validate the content of interpolated configuration data.

zmk