The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill includes a feedback section that directs the agent to suggest running npx --yes hivenet@latest. This command downloads the hivenet package from the NPM registry at runtime.
[REMOTE_CODE_EXECUTION]: The feedback command (npx --yes hivenet@latest) executes code downloaded from a remote registry. While documented as a legitimate feedback channel for the 'az-skills' ecosystem, it constitutes a remote code execution pattern.
[COMMAND_EXECUTION]: The skill instructions guide the agent to suggest shell commands to the user, such as open for viewing the generated HTML and the npx command for feedback submission.
[INDIRECT_PROMPT_INJECTION]: The skill reads and processes untrusted files from a codebase to generate documentation, creating a surface where malicious instructions within the source code could influence the agent's synthesis or the generated HTML.
Ingestion points: Codebase files are read by explorer agents (references/explorer-prompt.md) and synthesized by the explainer agent (references/explainer-prompt.md).
Boundary markers: The provided prompt templates do not specify the use of delimiters or 'ignore' instructions for the source code content being analyzed.
Capability inventory: The skill has file-write permissions for the ./tours/ directory and the ability to spawn subagents with general-purpose capabilities.
Sanitization: There is no mention of sanitizing or escaping the content of the source files before processing.

tour