pw-aippt-old

Pass

Audited by Gen Agent Trust Hub on Sep 25, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external URLs or text blocks to generate slide content and image prompts.
  • Ingestion points: The /pw-aippt command described in SKILL.md accepts external URLs or text as input.
  • Boundary markers: The prompt generation process uses section markers such as 【风格约束】, 【不可改区域】, and 【可改区域】 to guide the model, but these do not prevent adversarial instructions embedded in the source content.
  • Capability inventory: The skill uses subprocess calls to pdftoppm in scripts/pdf-to-images.js and performs file system operations to manage project directories, prompt files, and images.
  • Sanitization: There is no documented sanitization or filtering of the input content before it is interpolated into instructions for the agent.
  • [COMMAND_EXECUTION]: The script scripts/pdf-to-images.js executes system commands using child_process.execSync. Specifically, it invokes pdftoppm to convert PDF files into PNG images. While the file paths are quoted, this pattern allows for local system interaction based on project file names.
  • [EXTERNAL_DOWNLOADS]: The documentation in SKILL.md and references/01_导出方法.md guides users to install the poppler library via system package managers like Homebrew or APT. This is an expected dependency for the stated purpose of PDF processing and is documented clearly.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 25, 2026, 04:03 AM