The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill contains instructions to download and execute installation scripts for the Veris CLI tool directly from the vendor's official GitHub repository if the binary is missing.
Evidence: curl -LsSf https://raw.githubusercontent.com/veris-ai/veris-cli/main/scripts/install.sh | sh in SKILL.md.
Evidence: powershell -c "irm https://raw.githubusercontent.com/veris-ai/veris-cli/main/scripts/install.ps1 | iex" in SKILL.md.
[COMMAND_EXECUTION]: The skill performs various system operations to inspect the environment, manage containers, and handle credentials.
Evidence: Use of git rev-parse, docker image inspect, mktemp -d, and umask 077 to secure the execution environment.
[INDIRECT_PROMPT_INJECTION]: The skill reads application source code to identify SDK imports and vendor hostnames, creating an attack surface where malicious repository content could influence agent behavior.
Ingestion points: Local application source code files and remote twin manual markdown files.
Boundary markers: The skill instructs the agent to focus on specific code structures (e.g., imports, base-URL settings) but lacks explicit delimiters to isolate untrusted code content.
Capability inventory: The skill utilizes the veris CLI tool and has permissions for filesystem writes and container management.
Sanitization: Employs strict permission settings (umask 077) for temporary directories and uses jq to inspect metadata keys instead of raw data to prevent accidental disclosure.
[EXTERNAL_DOWNLOADS]: The skill downloads necessary configuration, tools, and scripts from the vendor's GitHub infrastructure.
Evidence: Fetches resources from raw.githubusercontent.com/veris-ai/.
[DYNAMIC_EXECUTION]: The skill generates build artifacts and configuration files at runtime to define the execution environment.
Evidence: Instructions to create a Dockerfile.veris for custom test images and the dynamic generation of the .veris/twin.yaml configuration file.

setup