book-writing-workspace
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes
subprocess.runwithintemplates/scripts/build_review_pdf.pyto orchestrate the PDF build process. It specifically invokes Docker to run thevvakame/reviewimage, which is a standard, well-known community tool for Re:VIEW-based publishing. This execution is scoped to the build pipeline and does not accept unvalidated external command strings. - [INDIRECT_PROMPT_INJECTION]: The writing and reviewer agents are designed to ingest and process Markdown manuscript files. While this creates a surface for indirect prompt injection where malicious instructions could be embedded in the text being reviewed, the skill includes explicit instructions for the agents to follow specific review criteria (P1/P2/P3) and writing guidelines, which serves as a baseline for safe operation.
- [EXTERNAL_DOWNLOADS]: The build process involves pulling and running a Docker image (
vvakame/review). This is a legitimate dependency for the Re:VIEW publishing toolchain and is clearly documented within the skill's instructions and dependencies list.
Audit Metadata