product-marketing
Pass
Audited by Gen Agent Trust Hub on Jul 21, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The 'Auto-draft from codebase' feature reads untrusted files within the repository (README, landing pages, marketing copy, package.json) to generate content. This creates a vulnerability surface for indirect prompt injection where malicious instructions hidden in these files could influence the agent's behavior during the drafting process.
- Ingestion points: scans codebase files (README.md, landing pages, package.json, etc.) in
SKILL.mdStep 2. - Boundary markers: No explicit delimiters are used when processing these files.
- Capability inventory: The skill is limited to reading files and writing to a specific location (
.agents/product-marketing.md). It does not have network access or arbitrary command execution capabilities. - Sanitization: The instructions require the agent to present the draft to the user for review and correction before saving, which serves as a human-in-the-loop security control.
Audit Metadata