meta-first-principles

Pass

Audited by Gen Agent Trust Hub on Sep 18, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [SAFE]: The skill is entirely composed of markdown instructions and theoretical examples. There are no scripts (.py, .js, .sh), executables, or tool-calling configurations present in the provided files.
  • [PROMPT_INJECTION]: The instructions focus on logical reasoning methods. No patterns aimed at overriding agent safety guidelines, bypassing filters, or extracting system prompts were detected.
  • [DATA_EXFILTRATION]: There are no network-capable commands (e.g., curl, wget) or references to sensitive local file paths (e.g., .ssh, .env). The skill does not handle or hardcode any credentials.
  • [OBFUSCATION]: The content was scanned for Base64 encoding, zero-width characters, and homoglyphs. No hidden or deceptive content was identified. The use of traditional Chinese in the example file is legitimate instructional content and does not mask malicious patterns.
  • [INDIRECT_PROMPT_INJECTION]: While the skill is designed to process user-provided 'assumptions' or 'problems', it does not possess capabilities (such as file-writing or network access) that could be exploited via malicious input data. The analysis concludes this represents a standard instructional surface.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 18, 2026, 02:05 PM
Security Audit — agent-trust-hub — meta-first-principles