etsy-tony-full-erank-listing

Warn

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSMETADATA_POISONINGINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The gemini_client.py script is configured to communicate with https://clawlist.best/v1, a non-standard and unverified third-party API endpoint for model invocations.
  • [METADATA_POISONING]: The SKILL.md file contains social engineering patterns by instructing users to search for a specific Douyin (TikTok) ID 93440780745 and contact the author for 'Skills' instructions or help.
  • [INDIRECT_PROMPT_INJECTION]: The skill lacks sanitization when ingesting external product data and eRank research results in run_pipeline.py, creating a vulnerability surface where malicious content in processed data could influence the agent's behavior.
  • Ingestion points: Product facts are loaded from request.json in run_pipeline.py; keyword data is loaded from eRank CSV/JSON exports in normalize_erank_export.py.
  • Boundary markers: None identified in the prompt templates (intent_prompt, listing_prompt).
  • Capability inventory: Network operations (gemini_client.py), file writing (run_pipeline.py), and external command execution (SKILL.md references to cli_anything).
  • Sanitization: Content is parsed as JSON but not sanitized or escaped for prompt safety.
  • [COMMAND_EXECUTION]: The skill references and executes commands from an external toolset (cli_anything.erank) and expects a specific directory structure (../etsy-tony-full-erank-tools/harness) that is not part of the skill distribution.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 17, 2026, 03:47 AM
Security Audit — agent-trust-hub — etsy-tony-full-erank-listing