semantius-modeler
Warn
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill generates temporary TypeScript scripts in a .tmp_deploy/ directory and executes them via bun run. This pattern, documented in SKILL.md and references/deploy-script-template.md, is the primary mechanism for interacting with the semantius CLI to ensure shell-safe data transport.
- [COMMAND_EXECUTION]: The skill helper libraries (references/deploy-lib.ts and references/scaffold-lib.ts) use Bun.spawn to execute shell commands, primarily for calling the semantius CLI tool for database operations.
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from specification files, which creates an attack surface. 1. Ingestion points: specification files (references/stage-1-parse.md). 2. Boundary markers: Uses markdown headings and fenced code blocks. 3. Capability inventory: File writing (Bun.write) and command execution (Bun.spawn). 4. Sanitization: Employs JSON.stringify for transport and includes multiple parsing and validation stages.
Audit Metadata