The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The launcher script downloads dependencies from the author's GitHub organization (getcolors). These components (red, package-neon-multi-node-red, colors-compute-red) are pinned to specific Git commits for integrity and are managed via a local cache directory.\n- [REMOTE_CODE_EXECUTION]: The red launcher script performs remote code execution by using 'bun install' to fetch packages from GitHub and subsequently loading them into the runtime environment.\n- [DYNAMIC_EXECUTION]: The skill employs dynamic imports to load its core logic from a computed cache path. It also supports the NEON_MULTI_NODE_LIB_ROOT environment variable, which allows the dynamic loading of arbitrary code from a specified local directory.\n- [PRIVILEGE_ESCALATION]: The instructions in SKILL.md guide the user to use 'sudo' to retrieve sensitive information, such as generated passwords stored in the /etc/neon/secrets/ protected directory.\n- [INDIRECT_PROMPT_INJECTION]: The skill has a vulnerability surface for indirect prompt injection, as it processes external state from 'colors.yml' and various environment variables to drive complex operations through powerful tools like Ansible and OpenTofu.\n
Ingestion points: Ingests configuration from the local 'colors.yml' file and overlays credentials from 'COLORS_PAR_*' environment variables.\n
Boundary markers: The skill does not explicitly describe the use of delimiters or 'ignore embedded instructions' markers when processing the desired state.\n
Capability inventory: The skill possesses capabilities for subprocess execution (Ansible, OpenTofu, AWS CLI, SSH, PSQL), network communication (AWS, Cloudflare), and file system writes.\n
Sanitization: The skill documentation notes the use of non-secret desired state, but there is no evidence of strict schema validation or sanitization for the inputs driving the operational DAG.

package-neon-multi-node-red