last30days

Pass

Audited by Gen Agent Trust Hub on Mar 18, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data from external social media and web sources.\n
  • Ingestion points: Untrusted content is ingested through scripts/lib/openai_reddit.py, scripts/lib/xai_x.py, and scripts/lib/websearch.py.\n
  • Boundary markers: The skill instructions do not mandate the use of delimiters or explicit "ignore instructions" warnings for the AI when it synthesizes the research results.\n
  • Capability inventory: The skill has the capability to make network requests (scripts/lib/http.py) and write files to the local system (scripts/lib/render.py).\n
  • Sanitization: The skill performs data cleaning, including text truncation and date validation, in scripts/lib/normalize.py.\n- [COMMAND_EXECUTION]: The skill executes the local Python script scripts/last30days.py to perform its research and processing tasks.\n- [DATA_EXFILTRATION]: The skill performs network requests to well-known and established services, including the OpenAI, xAI, and Reddit APIs, to fetch content relevant to the research topic.\n- [SAFE]: The skill demonstrates good security hygiene by instructing the user to apply restrictive file permissions (chmod 600) to the configuration file containing API keys.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 18, 2026, 05:44 AM