Skip to content

dcleveland0/promptworkshop

v0.5.6Proprietary

Harden rough coding tasks into scoped, testable instructions before execution.

GitHubOpen Website

PromptWorkshop for ChatGPT and Codex

PromptWorkshop turns rough coding requests into scoped, testable instructions before implementation begins.

Start here

  1. Install and enable the PromptWorkshop plugin.
  2. Authenticate when the MCP connection opens the PromptWorkshop sign-in flow.
  3. Ask: "Harden this task before I start coding: ."
  4. Review the returned prompt before authorizing implementation.

The plugin connects to https://www.promptworkshop.io/api/mcp/v1. It does not bundle credentials or require a local PromptWorkshop server. Task text and any attachments or repository brief you choose to include are sent to PromptWorkshop for processing under its privacy policy.

For an explicitly invoked /workshop-loop, the host may execute and audit the exact downstream result for up to five rounds. A full cycle can incur charges for up to five optimize runs and five Result Audits, plus host execution. Result Audit evaluates caller-reported output; it does not execute code or independently verify tests. The separate promptworkshop_loop_status MCP tool reads durable Loop state when given a Loop ID; this host-managed workflow does not create that ID.

Manage suggestion levels — Off, Complex, Average, or Basic — in PromptWorkshop External API settings. When a host activates the PromptWorkshop skill for an ordinary coding task, it can check this preference and ask before starting a run. Activation depends on the platform; no task text is sent before you say yes.