objectstack-platform
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill content is educational and technical in nature, providing documentation for a software framework without any malicious instructions or hidden payloads.
- [COMMAND_EXECUTION]: The documentation describes various CLI operations using the 'os' binary (e.g., 'os dev', 'os build', 'os migrate', 'os validate'). These are standard administrative and development tools for the ObjectStack platform.
- [EXTERNAL_DOWNLOADS]: The skill references several vendor-scoped packages from the '@objectstack/*' organization and well-known dependencies like 'typescript' and 'vitest'. It also describes a package installation mechanism ('os package install') designed to fetch versioned artifacts from a catalog.
- [INDIRECT_PROMPT_INJECTION]: The platform's architecture involves ingesting metadata from configuration files ('objectstack.config.ts') and remote artifacts. While this technically creates a surface for indirect prompt injection if an agent were to process untrusted or maliciously crafted metadata, the documentation itself promotes secure practices like 'strict' validation and structural integrity checks.
- Ingestion points: Metadata files (SKILL.md), JSON artifacts (references/operations.md).
- Boundary markers: The platform uses Zod schemas for strict structural validation (references/_index.md).
- Capability inventory: Subprocess calls for CLI commands, file system writes for artifacts, and network operations for package management (SKILL.md, references/operations.md).
- Sanitization: Validation gates ('os validate') are built into the workflow to catch semantic and structural errors before execution (references/operations.md).
Audit Metadata