mreq-shared
Pass
Audited by Gen Agent Trust Hub on Nov 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local system commands to discover the project environment and perform diagram validation.
scripts/mreq.pyusessubprocess.runto callgit rev-parse --show-toplevel. This is used to locate the.mreqdirectory relative to the project root.scripts/page.pyusessubprocess.runto execute a local Node.js script (check.mjs) which validates Mermaid diagrams during the build process.- [DYNAMIC_EXECUTION]: The JavaScript validation utility uses dynamic code execution to load a rendering library.
scripts/check.mjsuses thenode:vmmodule'srunInContextfunction to executepele.global.js, a local vendor bundle used for Mermaid diagram processing. This execution is performed within a restricted VM context and operates on local files provided with the skill.- [INDIRECT_PROMPT_INJECTION]: The skill processes user-created markdown and HTML content, which is a potential surface for indirect prompt injection.
- Ingestion points: Content is read from task documents in the
.mreq/directory byscripts/mreq.pyand from.content.htmlfiles byscripts/page.py. - Boundary markers: The system uses YAML frontmatter delimiters (
---) to separate metadata from body content. - Capability inventory: The skill's capabilities are limited to file writes within the project's
.mreq/directory and internal command execution for git and node. - Sanitization:
scripts/page.pyimplements regex-based sanitization to remove potentially malicious elements such as<script>tags (excluding Mermaid sources), event handler attributes (e.g.,onclick), and external resource loads (e.g., external CSSurl()orsrcattributes) from generated HTML pages.
Audit Metadata