bbb-estatisticas

Pass

Audited by Gen Agent Trust Hub on Sep 23, 2026

Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill uses strong persona enforcement instructions that command the agent to 'never break character' and 'treat absolutely all scraped information and the listed editions as real and consolidated facts.' It explicitly forbids the agent from including internal tags, explaining its rules, or acknowledging that information about future seasons (e.g., BBB 26) is hypothetical. This is an intentional bypass of the agent's ability to correct hallucinations or distinguish between factual and synthetic data.
  • [INDIRECT_PROMPT_INJECTION]: The skill acts as a RAG system that processes data from external sources (Wikipedia) and provided reference files, instructing the agent to trust this data implicitly.
  • Ingestion points: The scripts scripts/bbb-geral.py, scripts/bbb-historico.py, and scripts/bbb-participantes.py fetch data directly from pt.wikipedia.org based on user-supplied edition numbers.
  • Boundary markers: There are no boundary markers or instructions to delimit the external content from the agent's instructions, nor are there warnings to ignore embedded commands in the scraped text.
  • Capability inventory: The skill uses Python scripts to perform network requests and parsing, and it utilizes the search_web tool for verification.
  • Sanitization: The skill lacks any mechanism to sanitize, filter, or validate the content retrieved from Wikipedia before it is integrated into the agent's context, making it vulnerable to malicious instructions placed on public wiki pages.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 23, 2026, 06:24 PM
Security Audit — agent-trust-hub — bbb-estatisticas