The Agent Skills Directory

[Dynamic Execution] (MEDIUM): The skill uses dynamic loading patterns to register commands and cogs. In the Node.js example, it uses fs.readdirSync and require(filePath) to load modules from the ./commands and ./events directories. In the Python example, it uses os.listdir and bot.load_extension to load modules from ./cogs. While these are standard patterns in bot development, they involve executing code from computed filesystem paths.
[Indirect Prompt Injection] (LOW): The templates provide ingestion points for untrusted data through Discord interactions and slash command options (e.g., the msg parameter in the greet command).
Ingestion points: interaction object (src/index.js) and ctx: discord.ApplicationContext (main.py).
Boundary markers: Absent; there are no delimiters or instructions to ignore embedded commands in the user-provided strings.
Capability inventory: The provided code examples are restricted to display/response actions (interaction.reply, ctx.respond), placing them in the LOW capability tier.
Sanitization: The templates do not include logic to escape or sanitize the msg or message inputs before they are interpolated into responses.

discord-bot-architect