Verification · Code review · LLM application security review
Static review of LLM call sites
Check how untrusted text reaches prompts and how outputs reach tools or HTML.
How to build it
- List every call site that builds a prompt from outside input.
- Mark each place where untrusted text enters a prompt and each place where generated output reaches a tool, a shell or HTML.
- Review each marked site on every change that touches it.
Signals in a repository
- Prompt template files or prompt-building functions with a named owner.
- Code paths where generated output reaches an exec, shell or innerHTML call with a sanitizer step between them.
- A pull request template item that asks about new prompt inputs.
In practice
No scanned repository runs a tool for this check. The steps above are the only guidance.