database-architect

Pass

Audited by Gen Agent Trust Hub on Mar 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the Bash tool to implement a 'Memory Protocol'. This involves reading and writing to local files within the .claude/context/memory/ directory (learnings.md, issues.md, decisions.md) to persist state and configuration between sessions.
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface (Category 8). 1. Ingestion points: User-provided database requirements, entities, and access patterns described in SKILL.md. 2. Boundary markers: Not explicitly defined for user input interpolation. 3. Capability inventory: Includes Bash, Write, and Edit tools for managing schema files and migrations. 4. Sanitization: No explicit sanitization or validation of the input requirements is described in the provided files. While this represents a standard risk surface for database engineering agents, no malicious intent was identified.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 6, 2026, 09:10 AM