migration-architect
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits a vulnerability surface for indirect prompt injection due to its handling of untrusted external data.
- Ingestion points: The core scripts
scripts/migration_planner.py,scripts/compatibility_checker.py, andscripts/rollback_generator.pyingest untrusted data from user-provided JSON files (migration specifications and database schemas) via CLI arguments (--input,--before,--after). - Boundary markers: The skill's instructions in
SKILL.mdandREADME.mddo not utilize protective delimiters or explicit instructions to the AI agent to ignore potentially malicious natural language commands embedded within the processed JSON fields (e.g., in column comments or migration descriptions). - Capability inventory: The AI agent is instructed to execute these scripts, read their JSON/Text outputs, and summarize the findings for the user. This flow allows data from the input files to reach the agent's context.
- Sanitization: The provided Python scripts use standard
json.load()andjson.dumps()for processing, but do not implement specific sanitization, validation, or filtering of string values that are subsequently interpreted by the AI agent.
Audit Metadata