The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill guides the agent to interact with and execute project-specific files, which represents a potential attack surface for indirect prompt injection if those files contain malicious instructions.
Ingestion points: The agent is instructed to inspect deno.json, package.json, lockfiles, and project source code.
Capability inventory: The skill involves executing code via deno run and deno test, with instructions for handling network access, environment variables, and subprocesses.
Boundary markers: The skill does not define specific prompt delimiters for separating untrusted data from instructions.
Sanitization: The instructions explicitly advise to "Validate untrusted input" when handling HTTP services.
[SAFE]: The skill promotes security best practices by explicitly advising against the use of broad permission flags like --allow-all (-A) and instead recommends using scoped permissions (e.g., --allow-net=api.example.com) to restrict the runtime environment.
[SAFE]: All external references and documentation links target the official Deno documentation (docs.deno.com), which is a trusted and well-known service.

deno