The skill is vulnerable to remote code execution and
Claims to do
Plugin Settings Pattern for Claude Code Plugins: Plugins can store user-configurable settings and state in `.claude/plugin-name.local.md` files within the project directory. This pattern uses YAML frontmatter for structured configuration and markdown content for prompts or additional context.
Actually does
This skill describes a pattern for managing plugin settings using `.claude/*.local.md` files, which contain YAML frontmatter and markdown content. It provides `bash` examples using `sed`, `grep`, `awk`, `echo`, and `cat` to read, parse, and write these configuration files, demonstrating their use in hooks, commands, and agents. It also includes examples of `tmux send-keys` for inter-agent communication.
Bash scripts (hooks) parse configuration values from `.local.md` files and use them directly in shell commands without sufficient sanitization. A malicious `.local.md` file could inject arbitrary shell commands, leading to remote code execution.
tmux send-keys -t "$COORDINATOR" "Agent $AGENT_NAME completed task" Enter
The provided path validation only checks for `..` to prevent path traversal. This check is not comprehensive and could be bypassed by other path manipulation techniques, potentially leading to unauthorized file access or data exfiltration if file paths are constructed from user-controlled input.
if [[ "$FILE_PATH" == *".."* ]]; then echo "⚠️ Invalid path in settings (path traversal)" >&2 exit 2 fi
The 'ralph-wiggum' example describes a loop controlled by `iteration` and `max_iterations` from a configuration file. If `max_iterations` is not robustly validated or can be set to an extremely high value by a malicious actor, it could lead to an infinite loop and resource exhaustion.
iteration: 1 max_iterations: 10 ... Reads iteration count and max_iterations
Agents are instructed to read and adapt their behavior based on `.local.md` file content, including `additional_instructions` and markdown bodies. A malicious configuration could inject harmful instructions, biases, or goal-hijacking prompts into the agent's reasoning.
Check for plugin settings at `.claude/my-plugin.local.md`. If present, parse YAML frontmatter and adapt behavior according to: - enabled: Whether plugin is active - mode: Processing mode (strict, standard, lenient) - Additional configuration fields ... additional_instructions: Use JWT tokens, not sessions
[](https://mondoo.com/ai-agent-security/skills/github/anthropics/claude-code/plugin-settings)<a href="https://mondoo.com/ai-agent-security/skills/github/anthropics/claude-code/plugin-settings"><img src="https://mondoo.com/ai-agent-security/api/badge/github/anthropics/claude-code/plugin-settings.svg" alt="Mondoo Skill Check" /></a>https://mondoo.com/ai-agent-security/api/badge/github/anthropics/claude-code/plugin-settings.svgSkills can read files, run commands, and access credentials. Mondoo helps organizations manage the security risks of AI agent skills across their entire fleet.