The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to execute various commands in the user's local shell environment using an executable path determined from a configuration file.
Evidence: Use of <cliPath> for commands such as status, open, editors, test, run, and build within SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow where the agent reads local project files and tool outputs to determine subsequent shell commands, creating a surface for indirect influence.
Ingestion points: The agent is instructed to read <projectRoot>/Library/UnitySkills/cli_config.json, ~/.unity_skills/registry.json, ProjectVersion.txt, and XML/JSON outputs from the CLI tool itself.
Boundary markers: The skill does not define specific delimiters for isolating external file content from the agent's instructions, though it mandates checking an enabled: true flag for authorization.
Capability inventory: The skill utilizes shell command execution (via cliPath) and local file system access across all described operations in SKILL.md.
Sanitization: The instructions recommend using --format json to ensure structured data handling, but the logic relies on the integrity of the local cli_config.json and the specified binary.
[DATA_EXPOSURE]: The skill requires the agent to access local metadata files containing project paths, process IDs, and environment configurations.
Evidence: Reading project-specific and global registry files located at ~/.unity_skills/registry.json and Library/UnitySkills/cli_config.json.

unity-cli