The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a broad search command (find "$HOME" -maxdepth 8) to locate its own plugin directory and subsequently executes multiple scripts (ensure-deps.sh, extract_mermaid.js, resilient_diagram.js) found at that location. This dynamic loading of executable content from a path computed at runtime from a broad search space poses a risk of arbitrary command execution if an attacker can influence the file structure within the user's home directory.
[PROMPT_INJECTION]: The skill implements an indirect prompt injection surface by reading content from a user-provided codebase (Step 5) and interpolating that content into specialized instructions for the LLM (Step 7).
Ingestion points: Step 5 reads various code files (.py, .ts, .js, .sql, etc.) from the target path.
Boundary markers: No specific delimiters or "ignore instructions" warnings are applied to the code content before it is passed to the specialists.
Capability inventory: The skill possesses Write, Edit, and Bash capabilities, which are used in subsequent steps (Steps 8 and 9) to validate and render diagrams.
Sanitization: There is no evidence of sanitization or filtering of the code content to prevent embedded instructions from influencing the agent's behavior during the diagram generation or validation phases.

mermaid-architect