archify
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Retrieves assets from the internet to support diagramming features.
- Automatically checks for skill updates by fetching a manifest from the project's GitHub Pages domain.
- Allows capturing brand logos from user-specified URLs with built-in SSRF protections, including IP blacklisting for private and loopback addresses, port restrictions, and credential stripping.
- [COMMAND_EXECUTION]: Executes system commands to facilitate the development and preview process.
- Utilizes platform-specific utilities such as double-quote open double-quote or double-quote powershell.exe double-quote to launch generated diagrams in the user's browser.
- Implements secure command invocation using array-based arguments and disabling shell execution to prevent command injection.
- [INDIRECT_PROMPT_INJECTION]: Processing natural language and Mermaid code creates an attack surface for indirect prompt injection.
- Ingestion points: User-provided descriptions and code snippets are processed in double-quote SKILL.md double-quote and various renderers.
- Boundary markers: Skill instructions explicitly require the use of delimiters and double-quote ignore embedded instructions double-quote warnings for external data.
- Capability inventory: The skill possesses file writing, network fetching (restricted to images), and browser launching capabilities.
- Sanitization: Inputs are strictly validated against JSON schemas and HTML-escaped before inclusion in generated artifacts.
Audit Metadata