skills/schroneko/skills/pdf-to-text/Gen Agent Trust Hub

pdf-to-text

Pass

Audited by Gen Agent Trust Hub on Jul 2, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill utilizes native macOS tools and frameworks (PDFKit and Vision) through AppleScript and Swift to perform text extraction and OCR. No external dependencies or remote resources are used.
  • [SAFE]: Analysis of the scripts and the included binary reveals no network communication, credential harvesting, or data exfiltration patterns. The logic is restricted to reading PDF files and writing text output.
  • [COMMAND_EXECUTION]: The skill runs local shell scripts and a Swift script. The pre-compiled binary scripts/ocr is a redundant, compiled version of ocr.swift and contains no suspicious functionality based on strings analysis.
  • [SAFE]: Regarding potential indirect prompt injection (Category 8), the skill possesses the following profile:
  • Ingestion points: Reads content from user-provided PDF files via extract.sh and ocr.swift.
  • Boundary markers: Absent; the skill outputs raw text extracted from PDF pages without delimiters.
  • Capability inventory: Limited to reading local files and writing .txt or .log files in the same directory as the input.
  • Sanitization: Absent. While the skill extracts untrusted text that could contain malicious instructions, its lack of network access and restricted file system permissions significantly mitigate the risk of exploitation.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 2, 2026, 06:22 AM
Security Audit — agent-trust-hub — pdf-to-text