ad-creative-intelligence

Pass

Audited by Gen Agent Trust Hub on Mar 14, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection. It ingests untrusted text data (ad headlines, body copy, and descriptions) from Meta and Google Ad Libraries. If an attacker embeds malicious instructions within an ad, these could be executed or followed by the agent during the Phase 3 (Cluster & Analyze) or Phase 4 (Gap Analysis) steps.
  • Ingestion points: scrape_meta_ads.py and scrape_google_ads.py outputs (SKILL.md).
  • Boundary markers: None specified; ad content is processed directly for clustering and analysis.
  • Capability inventory: Python script execution and file writing to the clients/ directory.
  • Sanitization: No explicit sanitization or filtering of ad copy is described before the content is passed to the analysis phase.
  • [COMMAND_EXECUTION]: The skill uses python3 to execute local scripts (scrape_meta_ads.py and scrape_google_ads.py) located within subdirectories. These scripts take user-supplied domain names as arguments. While this is the intended workflow, it represents a command execution surface.
  • [SAFE]: The requirement for an APIFY_API_TOKEN is clearly documented as an environment variable and does not involve hardcoded credentials. The output paths are restricted to a specific client directory structure.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 14, 2026, 06:02 PM