seo-cluster
Pass
Audited by Gen Agent Trust Hub on Aug 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the web to architect SEO strategies and content clusters, creating a surface for indirect prompt injection.\n
- Ingestion points: External data is ingested via WebSearch (related searches, PAA questions) and DataForSEO (SERP URLs) as described in SKILL.md (Steps 1 & 2) and references/serp-overlap-methodology.md.\n
- Boundary markers: The skill uses keyword normalization and structured JSON schemas to handle data, but lacks explicit prompt delimiters to separate external content from agent instructions in the final planning steps.\n
- Capability inventory: The skill has capabilities for network access (WebSearch, DataForSEO), file system writes (cluster-plan.json, cluster-briefs/), and cross-skill invocation (calling blog-write in references/execution-workflow.md).\n
- Sanitization: The skill utilizes a url_safety check within render_page.py for URL fetching and performs specific extraction of organic result components rather than raw page ingestion.\n- [COMMAND_EXECUTION]: The skill invokes local helper scripts render_page.py and dataforseo_costs.py using a tool runner (claude-seo run). This is used to safely render pages and check API costs.\n- [EXTERNAL_DOWNLOADS]: The documentation suggests the installation of the claude-blog skill from an external GitHub repository (github.com/AgriciDaniel/claude-blog) for automated content creation.
Audit Metadata