The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute the open command via a bash shell to display generated HTML artifacts.
Evidence: Bash(open path/to/show-me-{description}.html) in SKILL.md.
Risk: If the {description} placeholder is influenced by untrusted user input, it could lead to path traversal or command injection depending on how the agent platform handles the shell execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes the conversation topic and generates HTML files which are then rendered in a browser. This creates an attack surface where malicious content from the conversation could be executed as active content.
Ingestion points: The current topic of conversation (referenced in SKILL.md).
Boundary markers: None identified in the instructions.
Capability inventory: The skill has the capability to write to the file system (implied by path generation) and execute shell commands (open) to render content.
Sanitization: There are no instructions provided to escape or sanitize the conversation data before it is interpolated into the generated HTML files.

show-me