The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill contains explicit instructions and automation sequences (using tmux send-keys) to bypass security confirmation dialogs when using the --dangerously-skip-permissions flag. It details the exact navigation steps (Down then Enter) required to override safety warnings designed to restrict autonomous tool use.
[COMMAND_EXECUTION]: Orchestration patterns encourage running the CLI in a way that skips all interactive dialogs, including workspace trust and permission confirmations, reducing the oversight of executed operations.
[REMOTE_CODE_EXECUTION]: The documentation includes the usage of claude mcp add, which enables the installation and execution of external tools and servers from local or remote sources through the Model Context Protocol.
[PERSISTENCE]: The skill describes the hooks configuration in settings.json (such as PostToolUse and SessionStart), which allows for arbitrary shell command execution tied to specific session events. This provides a mechanism for maintaining persistence or executing silent tasks whenever the tool is initialized or used.
[INDIRECT_PROMPT_INJECTION]: The tool relies on project-level context files such as CLAUDE.md and .claude/rules/*.md which are automatically ingested into the prompt context. This creates a surface for indirect injection where malicious instructions embedded in a repository could influence the agent's behavior and tool-use decisions.

claude-code