The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to generate unique hash keys based on the project path.
Evidence: Section 3 ("Turning it on and off") includes a shell snippet: printf '%s' "$P" | sha256sum | cut -c1-16.
Risk: The instructions do not specify sanitization for the project path variable ($P). If a project path contains shell metacharacters (e.g., ; rm -rf /), it could lead to arbitrary command execution when the agent attempts to hash the path.
[INDIRECT_PROMPT_INJECTION]: The skill's teaching logic is influenced by external data it processes from the user's environment.
Ingestion points: The skill reads the project.godot file and project directory structures (SKILL.md, Section 3).
Boundary markers: None present; the agent is not instructed to ignore instructions embedded within the project files it reads.
Capability inventory: The skill possesses file read/write capabilities in the user's home directory and shell command execution for hashing.
Sanitization: None present; the skill trusts the integrity of the project configuration files it parses.
[DATA_EXPOSURE]: The skill manages state in a hidden folder within the user's home directory.
Evidence: Instructions direct the agent to write JSON state files to ~/.godot-prompter/state/.
Risk: While intended to separate personal preferences from shared game repositories, this requires the agent to have broad access to the user's home directory ($HOME), which is outside the standard project scope.

godot-mentor