check-reporting

Pass

Audited by Gen Agent Trust Hub on May 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [SAFE]: No indicators of malicious behavior, multi-layer obfuscation, or data exfiltration attempts were found in the skill's instructions or checklists.\n- [COMMAND_EXECUTION]: The skill provides a local Python utility (scripts/check_prisma_figure.py) to verify the mathematical consistency of PRISMA flow diagrams. This script operates deterministically using standard libraries (re, pathlib, json) and lacks capabilities for network access or interaction with sensitive system files.\n- [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it is designed to analyze untrusted manuscript content. \n
  • Ingestion points: Reads the manuscript file (e.g., manuscript/manuscript.md). \n
  • Boundary markers: None explicitly defined in the instructions to separate untrusted manuscript text from analysis prompts. \n
  • Capability inventory: Uses Bash, Write, and Edit tools. \n
  • Sanitization: None described beyond standard LLM reasoning. This represents a low-severity inherent risk for document-processing agents.
Audit Metadata
Risk Level
SAFE
Analyzed
May 5, 2026, 12:29 PM
Security Audit — agent-trust-hub — check-reporting