verify-bundle-sync

Pass

Audited by Gen Agent Trust Hub on May 19, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill executes local, read-only comparisons using standard CLI tools like jq and diff to ensure project directory integrity.
  • [DATA_EXFILTRATION]: The skill reads from .claude-plugin/marketplace.json and accesses the skills/ and plugins/ directories to perform its checks. This data is not transmitted externally and is handled within the local project context.
  • [PROMPT_INJECTION]: The skill uses skill names retrieved from the local configuration to build paths for comparison. This represents an indirect injection surface, but the risk is assessed as safe given the tool's limited read-only capabilities and the local origin of the data.
  • Ingestion points: .claude-plugin/marketplace.json (referenced in Step 1).
  • Boundary markers: None present for the ingested configuration data.
  • Capability inventory: Utilizes jq, diff, and test in Step 1 and 2 for local file system verification.
  • Sanitization: No explicit sanitization of skill names before path construction.
Audit Metadata
Risk Level
SAFE
Analyzed
May 19, 2026, 03:57 PM
Security Audit — agent-trust-hub — verify-bundle-sync