test-error-states
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions to fetch documentation and capability references from the vendor's domain (docs.reticle.sh) using 'curl'. It also uses 'npx' to download and run the vendor-maintained package '@reticlehq/server'.
- [COMMAND_EXECUTION]: The instructions include command-line examples for starting Google Chrome with remote debugging enabled and initializing a Model Context Protocol (MCP) server using environment variables and 'npx'. These are standard development operations for the skill's stated purpose.
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection because it is designed to ingest and process external data, specifically browser console logs and UI element content (e.g., error banners).
- Ingestion points: Browser console levels and UI text queries (SKILL.md).
- Boundary markers: None explicitly defined in the provided instructions.
- Capability inventory: Uses 'reticle_run', 'reticle_act_and_wait', and 'reticle_clock' to manipulate the environment and interact with the browser.
- Sanitization: No specific sanitization or escaping of the ingested console or UI data is mentioned.
Audit Metadata