The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute wmill CLI commands including script preview, script run, generate-metadata, and sync push. These commands allow the agent to run local code, interact with the workspace, and modify local configuration files like .script.lock and wmill-lock.yaml.
[DYNAMIC_EXECUTION]: The skill is centered around writing TypeScript code that is then executed via the Bun runtime. It describes patterns for defining a main function and utilizing the windmill-client SDK to dynamically run other scripts (runScriptByPath) or flows (runFlow) within the platform.
[INDIRECT_PROMPT_INJECTION]: The agent is prompted to read project-specific files such as AGENTS.md, rt.d.ts, and local script files to infer correct command arguments and metadata updates.
Ingestion points: Reads script files, rt.d.ts (resource types), and AGENTS.md (project preferences).
Boundary markers: None explicitly defined for script argument generation.
Capability inventory: Execution of arbitrary TypeScript code via wmill script preview and local file modification via wmill generate-metadata.
Sanitization: Not mentioned; the skill relies on the agent to pick 'plausible args' from the script's signature.
[DATA_EXPOSURE]: The skill documentation references the use of sensitive environment variables like WM_TOKEN for authentication and the RT (Resource Type) namespace for accessing secrets like API keys. These are standard platform features for the Windmill environment.

write-script-bun