thurview-publish
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads and executes the Cloudflare Wrangler utility via npx at a pinned version (4.147.0). This version could not be verified against official release records for the wrangler package, which may indicate a dependency hallucination or a supply-chain discrepancy.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests Git remote URLs from the repository being processed. This data is used by the check-scope.mjs script to authorize deployment targets. This ingestion of untrusted repository metadata represents an indirect prompt injection surface.
- [DYNAMIC_EXECUTION]: The skill generates and runs temporary bash scripts and utilizes node -e for runtime calculations. This dynamic approach is functional for multi-provider deployments but expands the command execution surface.
- [COMMAND_EXECUTION]: Uses system CLI tools (git, az, aws, gcloud) with the user's existing credentials. The skill follows best practices by not requesting or storing secret keys directly.
Audit Metadata