The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill instructs the agent to execute commands with root privileges to modify system-wide configuration.
Evidence: Use of sudo tee -a /etc/sysctl.conf to append kernel parameters and sudo sysctl -p to apply them in SKILL.md.
[COMMAND_EXECUTION]: The skill directs the agent to perform potentially destructive file operations and system queries.
Evidence: Recursive deletion commands rm -rf targeting specific application cache and index directories in the user's home folder (~/Library/Application Support/Cursor/Cache/, ~/.config/Cursor/Cache/, ~/.cursor/index/).
Evidence: Shell commands for resource monitoring and configuration checking (du -sh, cat /proc/sys/fs/inotify/max_user_watches).
[PERSISTENCE]: The skill implements persistent modifications to the host operating system's configuration.
Evidence: Changes made to /etc/sysctl.conf are designed to persist across system reboots.
[INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface by combining the ingestion of untrusted repository data with high-privilege execution capabilities.
Ingestion points: Processes workspace files, .cursorignore, and .gitignore content to diagnose indexing issues.
Boundary markers: None identified in the instructional flow.
Capability inventory: Includes full Bash access with sudo privileges and file deletion capabilities.
Sanitization: No specific sanitization or validation of the repository data is performed before it influences diagnostic actions.

cursor-indexing-issues