zettaranc

Warn

Audited by Gen Agent Trust Hub on Aug 12, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches trading system data and configuration from a third-party GitHub repository (chenleiitaz/zettaranc-knowledge) using the GitHub CLI tool.\n- [COMMAND_EXECUTION]: The skill instructs the agent to perform environment detection and data management using shell commands such as ls, gh, and base64, as seen in SKILL.md and references/05-ingest-workflow.md.\n- [PROMPT_INJECTION]: The skill implements an ingestion workflow for incorporating new trading materials into its knowledge base, which creates a surface for indirect prompt injection.\n
  • Ingestion points: The raw/01-zettaranc/ directory is used to collect untrusted files for processing (documented in references/05-ingest-workflow.md).\n
  • Boundary markers: There are no explicit delimiters or safety instructions defined to prevent the agent from obeying instructions embedded in the ingested content.\n
  • Capability inventory: The skill possesses the ability to read and write local files in the wiki/ directory and perform network operations via the gh tool.\n
  • Sanitization: The workflow described in references/05-ingest-workflow.md lacks content validation, filtering, or sanitization logic for ingested data.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 12, 2026, 07:54 AM
Security Audit — agent-trust-hub — zettaranc