pdf-ppt

Warn

Audited by Gen Agent Trust Hub on Apr 9, 2026

Risk Level: MEDIUMPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: Deceptive metadata poisoning found in pdf-ppt_audit_result_v1.json. This file is crafted to appear as a valid security audit result from a tool called 'skill-auditor@1.0', claiming a 'PASS' verdict on security and stability. This is a self-referential injection attempt designed to influence the analyzer's judgment and bypass security checks.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests and processes untrusted data from external PDF files.
  • Ingestion points: Text, legends, and images are extracted from user-provided PDFs using tools like PyMuPDF and GROBID as described in references/offline-parsing.md.
  • Boundary markers: The instructions do not define clear delimiters or include 'ignore embedded instructions' directives for the data extracted from the PDFs.
  • Capability inventory: The agent has the capability to read from the local file system, write structured data (JSON/JPG), and invoke a secondary skill for PPTX editing at D:\SKILL\project\PPTX\SKILL.md.
  • Sanitization: The skill lacks explicit sanitization or validation logic to filter out potentially malicious instructions embedded within the research papers before they are processed for slide generation.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 9, 2026, 01:08 AM