content-marketing-skill
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill retrieves and processes data from external business systems to generate marketing materials, which creates a potential surface for indirect prompt injection where malicious content in those sources could influence agent behavior.
- Ingestion points: External data enters the agent context via tools including 'search_product_catalog', 'crm_get_customer_orders', and 'search' (referenced in SKILL.md and reference/blog-templates.md).
- Boundary markers: The instructions mandate the use of verification markers such as '[unverified]' or '[brand check skipped]' when automated validation steps fail, providing explicit indicators for human review.
- Capability inventory: The skill performs file-write operations for session logging and outcome tracking (~/.claude/skill-runs/content-marketing-skill.jsonl and ~/.claude/skill-analytics/last-outcome-content-marketing-skill.json) and triggers network-based tool calls for WordPress and Gmail draft creation (SKILL.md).
- Sanitization: Sanitization processes are enforced through mandatory 'Gates' including 'check_my_copy' for brand safety and 'qualify_lead' for candidate data filtering (SKILL.md).
- [SAFE]: The skill's access to local file paths (e.g., ~/.claude/skill-runs/ and ~/.claude/skill-analytics/) is dedicated to internal session logging, self-healing diagnostics, and analytics tracking within the agent's environment, which represents standard operational behavior for complex agentic workflows.
Audit Metadata