The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to execute multiple shell commands to manage configuration, including cp -p to clone file permissions, jq for JSON manipulation, and chmod +x to make a launcher script executable within the user's $HOME/.claude directory.
[EXTERNAL_DOWNLOADS]: The troubleshooting documentation indicates that the skill's infrastructure fetches data from https://msu.io/builder/notices. This introduces a dependency on an external domain not listed as a trusted vendor.
[INDIRECT_PROMPT_INJECTION]: The skill establishes an ingestion point for remote data (notices from msu.io) that is rendered directly into the Claude Code status line.
Ingestion points: Remote notices fetched via the launcher script.
Boundary markers: None identified in the prompt instructions to delimit the remote content from agent instructions.
Capability inventory: The skill has access to shell execution (bash), file writes (mv, cp), and network tools (curl).
Sanitization: No explicit sanitization or filtering of the remote notice content is described before it is displayed to the user and potentially read by the agent.
[DYNAMIC_EXECUTION]: The skill modifies settings.json to instruct Claude Code to execute a local shell script (msu-statusline.sh) every time the status line is rendered. Furthermore, the launcher script is designed to dynamically execute a stored command from msu-statusline.prev using bash -c to preserve the user's original status line functionality.

msu-statusline-install