The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes several shell utilities including cp, mv, rm, rmdir, and jq to manage file operations and state transitions during the uninstallation process. It also invokes the claude CLI to programmatically uninstall the targeted plugin.
[DYNAMIC_EXECUTION]: In the verification step (step 5), the skill extracts a command string from the settings.json file and executes it using bash -c. This pattern allows for the execution of arbitrary code defined within a configuration file.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from local configuration and backup files (settings.json and msu-statusline.prev) to restore the user's previous environment state. This data flows into the command execution path during verification.
Ingestion points: Reads from settings.json and msu-statusline.prev within the Claude configuration directory (defaulting to ~/.claude).
Boundary markers: The skill does not employ delimiters or explicit safety instructions when handling the content of the restored command strings.
Capability inventory: The skill possesses file write/delete permissions and the ability to spawn shell processes through bash -c.
Sanitization: The skill performs a basic string check to ensure the restored command does not name the plugin being removed, but it lacks comprehensive validation or escaping for the shell commands it restores and executes.
[DATA_EXPOSURE]: The skill accesses and modifies the settings.json file in the user's Claude configuration directory. This file contains global agent settings, plugin configurations, and potentially sensitive environment-specific metadata.

msu-statusline-uninstall