product-marketing-context

Pass

Audited by Gen Agent Trust Hub on Feb 20, 2026

Risk Level: SAFE
Full Analysis
  • [Prompt Injection] (SAFE): No instructions were found that attempt to override AI safety guidelines or extract system prompts.
  • [Data Exposure & Exfiltration] (SAFE): No sensitive file access (like SSH keys) or network operations were detected. The skill only interacts with local project documentation.
  • [Indirect Prompt Injection] (LOW): The skill has an ingestion surface as it reads local repository files (README, package.json, etc.) to auto-draft content.
  • Ingestion points: Reads project files such as README, landing pages, and package.json in Step 2.
  • Boundary markers: Absent; there are no specific delimiters used to separate the ingested file content from the agent's instructions.
  • Capability inventory: The skill can read local project files and write to the .claude/ directory.
  • Sanitization: Absent; the skill does not explicitly sanitize or validate the content of the files it reads before processing them.
  • [Remote Code Execution] (SAFE): No patterns of downloading or executing remote code or packages were found.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 20, 2026, 06:05 AM