promptdeck
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions for using a local CLI tool
promptdeck. The tool includes flags such as-out,-out-dir, and-manifestwhich are designed to write or overwrite files and directories on the local filesystem. This is a core functional feature of the tool, and the skill includes explicit 'Safety Rules' warning the user that it will overwrite files without confirmation. - [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow where external data is ingested and processed. This creates a potential surface for indirect instructions to be introduced into the agent's context.
- Ingestion points: Untrusted data enters the agent context through the
-varsand-matrixJSON files processed by therenderandmatrixsubcommands. - Boundary markers: No explicit boundary markers or instructions to ignore embedded content are provided in the template rendering process.
- Capability inventory: The tool's capabilities are limited to local file system writes and stdout rendering; it does not have internal network or subprocess execution capabilities.
- Sanitization: The tool uses Go's
text/templatefor interpolation but does not perform sanitization or escaping of the variable content before rendering. - [EXTERNAL_DOWNLOADS]: The skill includes instructions to install the
promptdeckbinary via the Homebrew package manager or by cloning and building from the author's GitHub repository (github.com/itamaker/promptdeck-skill). These installation steps point to the vendor's own official distribution channels.
Audit Metadata