package-walter-green

Pass

Audited by Gen Agent Trust Hub on Oct 5, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCREDENTIALS_UNSAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The green launcher script downloads its core library logic and transitive dependencies from the vendor's official GitHub repositories, specifically github.com/getcolors/walter.git, github.com/getcolors/colors-compute.git, and github.com/getcolors/green.git.\n- [REMOTE_CODE_EXECUTION]: The skill uses Babashka's add-deps functionality to pull remote Git repositories into the runtime environment and execute the resulting code, which is a standard pattern for this launcher's architecture.\n- [CREDENTIALS_UNSAFE]: The skill contains logic to read sensitive credential files from the user's workstation (such as ~/.claude/.credentials.json, ~/.codex/auth.json, and ~/.pi/agent/auth.json) to synchronize them with the remote development machine as a documented environment-seeding feature.\n- [COMMAND_EXECUTION]: The skill manages remote environments by invoking system-level CLI tools including OpenTofu for infrastructure as code, Ansible for machine provisioning, and ssh for host connectivity and configuration management.\n- [INDIRECT_PROMPT_INJECTION]: \n
  • Ingestion points: The skill reads project configuration from a colors.yml file located in the user's project directory (SKILL.md).\n
  • Boundary markers: While it lacks explicit delimiters for the configuration data, the skill documentation provides "Non-negotiable safety rules" to prevent the AI from exposing secrets or overriding critical settings.\n
  • Capability inventory: The skill possesses extensive capabilities including infrastructure provisioning, file system writes, and remote command execution via OpenTofu and Ansible (SKILL.md).\n
  • Sanitization: The skill performs validation on configuration keys and explicitly refuses to build if placeholder values are detected in the configuration file (references/configuration.md).\n- [DYNAMIC_EXECUTION]: The launcher employs dynamic resolution and execution of Clojure functions at runtime (requiring-resolve), allowing it to adapt to different project configurations and library versions.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 5, 2026, 10:04 AM