Usage
pflow probe executes a single node for exploratory testing. By default it returns metadata and template paths rather than dumping raw output.
Parameters
name=value arguments become the node’s params; a value’s type is inferred (5 is a number, true a boolean, {"a": 1} an object). A ${VAR} in a param is filled from the environment or from pflow settings set-env before the node runs — except inside a body (a shell command, a code block), whose ${…} belongs to that language. A variable set in neither place refuses the probe with the commands that set it. The node runs the way a workflow step does: a file path in a param that takes content (command, code, prompt, source, stdin, headers, output_schema) — one starting with ./ or ../, or one containing / and ending in .md, .txt, .py, .sh, .yaml, .yml or .json — is replaced by the file’s content, and params are validated as pflow run validates a step: an unknown param or an invalid param name is refused before the node runs (exit 2); a value the node itself rejects (an empty file_path, a bad timeout) is a node failure (exit 1). As in a run, the value of a sensitive-named param (api_key, auth_token, an env entry like API_TOKEN) and every pflow settings set-env value print as <REDACTED> wherever they appear.
Failures and exit codes
With
--output-format json, stdout always carries one document. A failure’s document has "success": false and an errors list (plus warnings); a successful one omits errors and carries warnings only when there are any.