shell

Pass

Audited by Gen Agent Trust Hub on Jun 18, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill implements a Retrieval-Augmented Generation (RAG) pipeline in references/vector-rag.md that is susceptible to indirect prompt injection. • Ingestion points: Data is retrieved from Couchbase collections using subdoc get text. • Boundary markers: There are no delimiters or 'ignore embedded instructions' warnings when passing the retrieved content to the ask command. • Capability inventory: The skill can perform cluster management, query execution, and file system operations (save) across various scripts. • Sanitization: No sanitization or escaping of the retrieved document content is performed before it is sent to the LLM.
  • [EXTERNAL_DOWNLOADS]: The skill uses curl to fetch information from the well-known service ifconfig.me in references/cluster-management.md. This is used to determine the local public IP for inclusion in the Couchbase Capella allowlist.
  • [COMMAND_EXECUTION]: The skill demonstrates dynamic command execution by piping the output of the query advise tool directly into the query command in references/cluster-management.md. This execution pattern relies on the integrity of the tool's output to prevent unintended query execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 18, 2026, 09:29 AM
Security Audit — agent-trust-hub — shell