keyword-wedge
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits a significant ingestion surface where untrusted data could influence agent behavior.
- Ingestion points: The skill reads local codebase content (Phase 2), organic search queries from Google Search Console (Phase 4), user behavior data from PostHog (Phase 4), and live web content from Google Keyword Planner via browser automation (Phase 5).
- Boundary markers: The agent prompts in
references/agent-prompts.mddo not currently specify strict delimiters or instructions to ignore potential commands embedded within the analyzed content (e.g., malicious strings in search queries or markdown files). - Capability inventory: The agent has the ability to perform network fetches, automate browser sessions, and write files to the local
.keyword-wedge/directory. - Sanitization: The skill does not explicitly define sanitization protocols for data retrieved from external APIs or browser scraping before it is processed or included in generated reports.
- [EXTERNAL_DOWNLOADS]: The skill connects to well-known external service endpoints to gather necessary SEO and analytics data.
- Fetches site performance data from Google's Search Console API and keyword metrics from Google Ads via browser automation.
- Retrieves user engagement and funnel data from the PostHog analytics platform API.
Audit Metadata