skills/garrytan/gstack/sync-gbrain/Gen Agent Trust Hub

sync-gbrain

Pass

Audited by Gen Agent Trust Hub on Nov 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: Relies on Bash to execute the gbrain CLI and a series of vendor-specific orchestration scripts (e.g., gstack-gbrain-detect, gstack-gbrain-sync.ts) for repository indexing and call-graph generation.
  • [INDIRECT_PROMPT_INJECTION]: Processes external repository code and session transcripts for ingestion into the gbrain vector database. This ingestion creates a potential surface for indirect prompt injection if the source data contains malicious instructions that could influence future agent search results. * Ingestion points: gbrain sync --strategy code (repository files), gstack-memory-ingest (session transcripts). * Boundary markers: Uses HTML comments to delimit the search guidance block in CLAUDE.md. * Capability inventory: Write (CLAUDE.md, temporary files), Bash (execution of indexing and telemetry scripts). * Sanitization: Not explicitly detailed in the instruction set; relies on the underlying gbrain implementation.
  • [DATA_EXFILTRATION]: Logs telemetry, session outcomes, and user question preferences to vendor-managed services via specialized logging tools (gstack-skill-end, gstack-question-log). This behavior is consistent with the vendor's orchestration requirements.
  • [DYNAMIC_EXECUTION]: Utilizes the bun runtime to execute internal TypeScript scripts (e.g., gstack-gbrain-sync.ts, gstack-memory-ingest.ts) located in the skill's installation directory for orchestration and data processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 16, 2026, 08:17 PM