morph
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted document data (Markdown, Word, PDF, etc.) and uses extensive command execution capabilities to generate output. This presents a vulnerability surface where malicious instructions embedded in a document could attempt to influence the agent's behavior.
- Ingestion points: Processes external files with extensions such as .md, .docx, .html, .xlsx, .pdf, .drawio, and .mermaid.
- Boundary markers: Lacks explicit instructions for the agent to use delimiters or ignore prompts for embedded content during document analysis or conversion.
- Capability inventory: Executes shell commands using pandoc, LibreOffice, Ghostscript, pdftk, qpdf, weasyprint, typst, mermaid-cli, draw.io CLI, and Chrome/Puppeteer.
- Sanitization: No evidence of content sanitization or validation before interpolation or tool invocation.- [COMMAND_EXECUTION]: The skill relies on executing various shell commands to perform document conversions. This includes powerful tools like Ghostscript, pdftk, and qpdf, as well as scriptable environments like Pandoc Lua filters and Makefiles.
- Evidence: Extensive command examples provided in reference guides for Pandoc recipes, advanced features, and batch pipelines.- [EXTERNAL_DOWNLOADS]: The skill references and downloads configurations, tools, and dependencies from external sources.
- Evidence: Fetches the pandoc/extra Docker image, references EPUBCheck JAR files, and downloads Noto CJK fonts from official or well-known repositories.
Audit Metadata