mreq-how
Pass
Audited by Gen Agent Trust Hub on Nov 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The
scripts/page.pyscript executes the systemnodeprocess to validate Mermaid diagrams. The execution is controlled, using static paths to internal scripts and a fixed list of arguments, which prevents command injection. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content by reading local source code and documentation files to generate its reports. This creates an attack surface for indirect prompt injection if malicious instructions are embedded in the analyzed codebase.
- Ingestion points: Local files such as README.md, AGENTS.md, and project source code files accessed during the 'Explore' step.
- Boundary markers: The skill uses a structured workflow but does not specify delimiters to separate untrusted codebase content from its internal reasoning instructions.
- Capability inventory: The skill can read local project files, write HTML documentation to the
.mreq/how/directory, and execute its internal Python-based page assembly script. - Sanitization: The
scripts/page.pyscript implements robust sanitization for the generated documentation. It escapes header values and bans dangerous HTML tags such as<script>,<meta>, and event handler attributes. Furthermore, it generates a Content Security Policy (CSP) with script hashes to ensure that only authorized, bundled scripts can execute within the generated offline page.
Audit Metadata