publishing

Warn

Audited by Gen Agent Trust Hub on Nov 14, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill instructs the agent to execute a Node.js script located at a relative path (../story-maintenance/scripts/story.js) or a user-provided directory (<checkout>/bin/story.js) if the main CLI tool is not installed. This pattern of executing code from computed or untrusted paths poses a risk if the environment or input is not properly validated.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-controlled project files, creating a vulnerability to indirect prompt injection.
  • Ingestion points: story.md (specifically description, keywords, and ai-disclosure fields), submission/blurb.md, and matter pages containing quoted material.
  • Boundary markers: The skill does not define boundary markers or specify that the agent should ignore instructions found within these files.
  • Capability inventory: The skill can execute shell commands, perform file system operations (writes), and build project assets.
  • Sanitization: No sanitization or validation of the ingested text is performed before it is used to influence the agent's behavior or project outputs.
  • [COMMAND_EXECUTION]: The skill relies on several external command-line utilities to perform its core functions.
  • Evidence: Frequent use of the story CLI, as well as java -jar epubcheck.jar, ace, weasyprint, prince, and pagedjs-cli for validation and rendering.
  • [EXTERNAL_DOWNLOADS]: The skill references and recommends several third-party tools for ebook and PDF production.
  • Evidence: Mentions of EPUBCheck, Ace by DAISY, Prince, WeasyPrint, and Paged.js. These are established tools in the publishing domain.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 14, 2026, 07:26 PM