verifying-external-behavior
Pass
Audited by Gen Agent Trust Hub on Aug 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides numerous examples of shell commands for verifying system behavior. It uses uv to run arbitrary Python code snippets, curl to query remote APIs and inspect response headers/status codes, and docker to instantiate throwaway service containers for testing integrations.
- [EXTERNAL_DOWNLOADS]: The skill instructions involve downloading external software and data. Specifically, it demonstrates using the uv tool to fetch and run Python libraries from public registries and docker to pull alpine-based service images from Docker Hub.
- [PROMPT_INJECTION]: The skill describes workflows that ingest data from untrusted external sources, creating an indirect prompt injection surface. 1. Ingestion points: Processing remote API responses (e.g., via curl), inspecting contents of build artifacts (unzip -l), and scraping external documents. 2. Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when processing this external content. 3. Capability inventory: The skill utilizes tools with network access (curl, requests), shell execution capabilities (uv, docker), and file system interaction (uv build). 4. Sanitization: No methods for sanitizing or validating external responses are mentioned in the provided examples or guidelines.
Audit Metadata