convex-authz
Pass
Audited by Gen Agent Trust Hub on Nov 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from user-provided source files located in the
convex/directory. These files serve as the input for the agent's regex-based scanning and subsequent refactoring. There is a potential risk that malicious instructions hidden in code comments or string literals within these files could attempt to manipulate the agent's behavior during the hardening process. No explicit boundary markers or sanitization steps are defined for the ingested file content. - [DYNAMIC_EXECUTION]: The skill performs automated code generation and refactoring. It is instructed to create new files (e.g.,
convex/model/auth.ts) and rewrite existing function handlers by injecting security checks (requireIdentity,requireOwner). While these modifications follow canonical templates, the generation process is driven by the structure of untrusted user code. - [COMMAND_EXECUTION]: The workflow includes the execution of
npx tsc --noEmitto verify the integrity of the applied fixes. This involves a local shell command execution within the project environment. Althoughtscis a standard and trusted development tool, executing commands based on a modified state of the project is a capability that should be monitored.
Audit Metadata