The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute system-level commands such as open, xdg-open, or start to launch the default web browser and display the generated HTML visualization.
[EXTERNAL_DOWNLOADS]: The skill requires the generated HTML to load external libraries (Mermaid and Tailwind CSS) from public CDNs at runtime. While these are well-known services, it represents a dependency on external code for the visualization to function correctly.
[PROMPT_INJECTION]: The skill instructions include a directive to bypass user confirmation ("Don't gate on permission: when it applies, produce the page as part of answering"), which increases the autonomy of the agent and reduces user oversight for file-writing and command-execution tasks. Additionally, the skill is susceptible to indirect prompt injection as it ingests untrusted data to generate the HTML content.
Ingestion points: User-provided code snippets or architectural descriptions processed in SKILL.md.
Boundary markers: Absent; the instructions do not specify using delimiters or "ignore instructions" wrappers for the data embedded in the HTML.
Capability inventory: The skill possesses the ability to write to the filesystem ($TMPDIR) and execute browser-opening commands.
Sanitization: Absent; there is no instruction to sanitize, escape, or validate external content before interpolating it into the Mermaid or HTML template.
[DATA_EXFILTRATION]: The skill writes generated content to the OS temporary directory ($TMPDIR or %TEMP%) and outputs the absolute path to the terminal. This behavior moves data from the agent's workspace into a global system path accessible to other processes.

visual-explainer