The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the find command to search the user's home directory for a specific directory structure. It then uses the result ($PLUGIN_DIR) as a base path to execute JavaScript files with node. This dynamic path resolution is susceptible to search-order hijacking if a malicious directory with the same name is placed in a location scanned by find.- [EXTERNAL_DOWNLOADS]: The skill's health check feature automatically invokes npm install to download and install the beautiful-mermaid package if it is missing. This introduces external code from the public NPM registry into the user's environment.- [REMOTE_CODE_EXECUTION]: Automatic installation of third-party packages from unverified sources (like beautiful-mermaid via NPM) creates a risk of remote code execution, as the package manager may run arbitrary installation scripts defined by the package author.- [PROMPT_INJECTION]: The skill ingests untrusted user input via the $ARGUMENTS variable and incorporates it into a JSON configuration file. Ingestion point: $ARGUMENTS in SKILL.md. Boundary markers: None. Capability inventory: Write to .claude/mermaid.json, Bash execution of dynamic paths. Sanitization: None explicitly defined. This creates a surface for schema confusion if the configuration file is later processed by other agent components.

mermaid-config