The Agent Skills Directory

[DYNAMIC_EXECUTION]: The diagnostic procedure uses the find command to search for an executable script (mcp-launcher.sh) within a plugin cache directory and then executes it via a dynamically computed path. This represents a risk where the agent might execute a script discovered at runtime rather than from a pre-defined path.
[REMOTE_CODE_EXECUTION]: The skill provides instructions to the user to install the Bun runtime using a piped-to-shell command (curl -fsSL https://bun.sh/install | bash). While this targets a well-known service, the installation pattern involves executing a remote script directly.
[COMMAND_EXECUTION]: The uninstall feature utilizes the Bash tool to perform several system-level actions, including the removal of plugin directories via rm -rf. While intended for cleanup, this demonstrates the skill's ability to perform significant file system modifications.
[INDIRECT_PROMPT_INJECTION]: The skill ingests results from external MCP tools and is commanded to display them verbatim, bypassing typical agent summarization. This creates a potential surface for indirect prompt injection if the tool output contains malicious instructions.
Ingestion points: Results from mcp__claude_buddy__* tools (found in SKILL.md).
Boundary markers: None; the skill explicitly forbids paraphrasing or altering the tool output.
Capability inventory: Access to the Bash tool for file system operations and command execution (found in SKILL.md).
Sanitization: None; raw tool output is relayed to the user.

buddy