The Agent Skills Directory

[DYNAMIC_EXECUTION]: The green launcher script dynamically loads and resolves the agent-network-doks and green libraries at runtime using Babashka's dependency management.\n
Evidence: The script uses babashka.deps/add-deps and requiring-resolve to execute functions from libraries not present on the initial classpath.\n- [EXTERNAL_DOWNLOADS]: The skill downloads source code from the author's repositories on GitHub during the bootstrap process.\n
Evidence: The green script references https://github.com/getcolors/agent-network-doks.git and https://github.com/getcolors/green.git for library resolution.\n- [INDIRECT_PROMPT_INJECTION]: The skill deploys an AI agent (Claude Code) intended to process external data through a NetBird gateway, which presents an attack surface for indirect prompt injection.\n
Ingestion points: Data entering the isolated agent pod through the NetBird SOCKS5 gateway.\n
Boundary markers: Infrastructure includes NetworkPolicy and model allowlists to limit reach, but does not specify prompt-level delimiters for data isolation.\n
Capability inventory: The deployment environment includes kubectl access and DigitalOcean API integration via the COLORS_PAR_DO_TOKEN.\n
Sanitization: No specific input sanitization for the agent's natural language processing is described.

package-agent-network-doks-green