The Agent Skills Directory

[COMMAND_EXECUTION]: The skill employs a dynamic discovery mechanism using the find command to locate a plugin directory within the user's home folder ($HOME) and subsequently executes a shell script from that path (bash "$PLUGIN_DIR/scripts/ensure-deps.sh"). This dynamic execution from a computed path allows for the execution of arbitrary code if a malicious directory structure is present in the current project or workspace.
[COMMAND_EXECUTION]: The skill executes the d2 command-line tool to validate and render diagrams. This involves running shell commands on output files that are generated from user-provided input.
[EXTERNAL_DOWNLOADS]: The invocation of ensure-deps.sh indicates that the skill may download or install external dependencies. The source and integrity of these dependencies are not defined within the skill file, posing a potential supply chain risk.
[PROMPT_INJECTION]: The skill ingests untrusted Mermaid diagram code from $ARGUMENTS or external files without explicit sanitization or boundary markers. This exposes the agent to indirect prompt injection, where instructions hidden within the diagrams could attempt to manipulate the agent's conversion logic or subsequent actions.

d2-convert