huashu-nuwa

Fail

Audited by Gen Agent Trust Hub on Nov 16, 2026

Risk Level: CRITICALREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: Automated security scans identified confirmed malware within the skill package. Multiple research files, including examples/trump-perspective/references/research/06-timeline.md and examples/trump-perspective/references/research/03-expression-dna.md, as well as files in the X-Mastery-Mentor perspective, triggered FileRepMalware detections. One file is specifically flagged as a suspected HttpRequest-inf payload.
  • [EXTERNAL_DOWNLOADS]: The skill's research datasets contain several URLs flagged as malicious by reputation scanners. These include botnet-linked documents on ijels.com and phishing sites on blog.auspiciousassociates.com and davidschwertfeger.com. Visiting or processing these during the skill's research phase poses a significant threat.
  • [DATA_EXFILTRATION]: The x-mastery-mentor sub-skill includes instructions to use the computer-use and claude-in-chrome tools to perform automated scraping of user profiles on X (Twitter). This retrieves engagement metrics, post history, and profile data, which could be exfiltrated for large-scale data harvesting without user oversight.
  • [INDIRECT_PROMPT_INJECTION]: The skill's architecture ingests and processes large, untrusted datasets from public web sources. The lack of sanitization or explicit boundary markers around this external content creates a substantial vulnerability to indirect prompt injection attacks if the source material contains hidden malicious instructions.
  • Ingestion points: Markdown files in examples/[persona]/references/research/.
  • Boundary markers: None; the skill relies on the agent interpreting raw markdown content.
  • Capability inventory: The skill uses WebSearch, bash, and python3 for data gathering and analysis.
  • Sanitization: No sanitization logic was found in the provided scripts or templates.
  • [COMMAND_EXECUTION]: The skill provides and executes shell scripts such as download_subtitles.sh, which uses the yt-dlp utility and accepts external URLs as arguments. Furthermore, the main SKILL.md file contains logic that executes git commands on the host machine to perform automated version checks.
Recommendations
  • CRITICAL: 1 infected file(s) detected - DO NOT USE
  • CRITICAL: 4 file(s) identified as malware by FileRep - DO NOT USE
  • AI detected serious security threats
  • Contains 6 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
CRITICAL
Analyzed
Nov 16, 2026, 02:01 AM