genviral

Pass

Audited by Gen Agent Trust Hub on Mar 14, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The scripts/update-skill.sh script downloads updated skill files from the author's official GitHub repository (github.com/fdarkaou/genviral-skill). Additionally, several scripts fetch content from external URLs provided by the user for analysis.
  • [REMOTE_CODE_EXECUTION]: The self-update mechanism in scripts/update-skill.sh fetches remote scripts and overwrites local files, which are then marked as executable. This allows for runtime modification of the skill's logic.
  • [COMMAND_EXECUTION]: The orchestration scripts (orchestrate.py, run_all.sh) and the main CLI wrapper (genviral.sh) use subprocess.run and system shells to execute various Python and Bash tasks as part of the content generation pipeline.
  • [PROMPT_INJECTION]: The skill contains an indirect prompt injection surface. It ingests untrusted data from external websites (landing pages via lp_analyzer.py and community sites like Reddit as described in meta-ads/campaign/references/stage-prompts.md). This data is subsequently processed by the agent to generate ad copy. This risk is partially mitigated by a mandatory manual review step ('HARD GATE') before any content is posted.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 14, 2026, 04:04 AM