convex-backup

Pass

Audited by Gen Agent Trust Hub on Oct 17, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to use official Convex CLI commands (npx convex export, npx convex import, npx convex deploy) for database operations. These tools are provided by the vendor for managing its platform and are used here for their intended purpose.
  • [DATA_EXPOSURE]: The workflow involves handling sensitive database snapshots. The skill correctly identifies these as sensitive data and includes explicit hygiene rules to delete local copies after use and prevent them from being committed to version control systems.
  • [INDIRECT_PROMPT_INJECTION]: The skill reads data from database tables to verify the success of restoration drills.
  • Ingestion points: Data is ingested through MCP tools like tables and runOneoffQuery during the assertion step (SKILL.md).
  • Boundary markers: No specific delimiters are used during the verification of restored data.
  • Capability inventory: Subprocess execution via npx for exports, imports, and deployments; file system access for snapshot management.
  • Sanitization: Verification is limited to row count comparisons and spot-checks of records, which is appropriate for the stated goal of backup validation.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 17, 2026, 01:02 AM