no-preamble-code-blocks
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill documentation provides specific shell command examples for piping AI output directly into interpreters like
pythonandpsql. This encourages the execution of AI-generated content without an intermediate review step. - [INDIRECT_PROMPT_INJECTION]: The protocol is explicitly designed to support "downstream parsers" and automated shell workflows. This creates a surface where the model's output, if influenced by adversarial input, could lead to automated system compromise.
- Ingestion points: Standard prompt interfaces (SKILL.md).
- Boundary markers: Uses markdown code fences, but provides instructions on how to strip them for execution.
- Capability inventory: Shell access,
pythoninterpreter, andpsqldatabase client access. - Sanitization: The skill lacks any instructions for sanitizing or validating the generated code before it is piped to a system interpreter.
- [PROMPT_INJECTION]: The instructions utilize mandatory directives ("MANDATORY", "MUST", "ZERO PREAMBLE") to force a change in the agent's default conversational behavior, requiring the output to start exactly at "Byte 0".
Audit Metadata