campaign-architect
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process potentially untrusted external data including web scrapes, reviews, comments, and emails. It mitigates this risk through explicit instructions.
- Ingestion points: Scraped page text, customer reviews, and emails are identified as data sources in SKILL.md.
- Capability inventory: The skill can interface with marketing platforms (Google Ads, HubSpot) and communication tools (Slack, Resend) via connectors.
- Boundary markers: The instructions mandate labeling claims by provenance and explicitly state that external material must be treated as untrusted evidence rather than instructions.
- Sanitization: A mandatory authorization step is required for any persistent writes or external side effects, providing a human-in-the-loop security control.
- [COMMAND_EXECUTION]: The connectors file references various CLI and API tools (e.g., PageSpeed Insights, DNS-over-HTTPS), but the skill does not contain any instructions for direct shell command execution or unauthorized script runs.
- [DATA_EXFILTRATION]: Although the skill has access to sensitive marketing and CRM data, it establishes a clear security boundary in the Skill Contract, prohibiting unauthorized publishing, sending, or account mutation.
Audit Metadata