The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the Bash tool to validate and render diagrams. The command mmdc -i <filename>.mmd relies on a filename string generated by the agent based on user-supplied content. If the agent fails to sanitize the filename, it could potentially include shell metacharacters (e.g., ;, &, |) leading to arbitrary command execution.
Evidence: SKILL.md Steps 4 and 6 detail the execution of mmdc in a bash environment.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests untrusted user content to generate shell command arguments.
Ingestion points: Natural language descriptions and source code provided by the user (as described in SKILL.md, Step 1: "Understand the Request").
Boundary markers: The instructions lack explicit boundary markers or warnings to ignore instructions embedded within user-provided content.
Capability inventory: The skill utilizes the Bash tool to execute the mmdc utility.
Sanitization: While the skill suggests using "kebab-case" for filenames, it does not provide comprehensive sanitization logic to prevent command injection or path traversal when constructing shell commands.

mermaid