aura-review
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests source code from the application's
srcdirectory as identified in Step 3 of SKILL.md and scripts/scan-aura-usage.ts. The agent is instructed to manually skim the JSX context to understand component behavior; this process lacks boundary markers or explicit sanitization, potentially exposing the agent to instructions embedded in code comments or strings. The skill operates with a broad capability set including Bash, Write, and WebFetch tools. - [COMMAND_EXECUTION]: As described in Step 1 of SKILL.md, the skill runs a local scanning script using
tsxornpx --yes tsx, the latter of which involves executing code fetched from the npm registry. - [EXTERNAL_DOWNLOADS]: The skill fetches design guidelines and component information from
https://docs.cognite.com/aura-design-systemandhttps://cognitedata.github.io/flows-community-components/using theWebFetchtool. These are official vendor resources related to the Aura design system. - [DYNAMIC_EXECUTION]: The scanning script in
scripts/scan-aura-usage.tsdynamically loads thetypescriptcompiler library from the audited project'snode_modulesdirectory using theNODE_PATHenvironment variable, as noted in the Step 1 instructions.
Audit Metadata