jed-zed/ccg-codex-workflow/ccg-gptpro-plan

CCG GPT Pro Plan

This is a Codex + Gemini + GPT Pro planning workflow.

Load and follow skills/ccg-gptpro-bridge/SKILL.md.

Behavior

• Treat the argument as a planning task or plan-review input.
• Run Gemini before GPT Pro using the bundled Gemini preview helper with --prompt-template plan.
• Follow the Gemini Gate Before GPT Pro from skills/ccg-gptpro-bridge/SKILL.md: require a real CCG_GEMINI_RESPONSE_FILE, read a non-empty Gemini response from it, stop and do not create a GPT Pro bridge session if it is missing or empty, and do not invent Gemini findings.
• Include Codex's planning context, the Gemini response file path, and a concise Gemini findings summary in the GPT Pro prompt.
• Include the original CCG 需求完整性评分（0-10） with 目标明确性（0-3）, 预期结果（0-3）, 边界范围（0-2）, 约束条件（0-2）, 总分：X/10, and 判定：>=7 继续；<7 停止并提出补充问题.
• Require GPT Pro to output a Planning Readiness Scorecard for plan maturity, ambiguity, risk, testability, and whether Codex should proceed.
• Build a single-round planning prompt by default.
• Expected manual questions: 1.
• Maximum manual questions: 2.
• Round 2 only for blocker re-check or revised plan comparison.
• Use scripts/gptpro_bridge.py --mode plan --detach-preview --open-preview --gemini-response-file <CCG_GEMINI_RESPONSE_FILE> --gemini-summary-file <summary-file>.
• Read the saved response file only after the user manually saves it.
• Summarize and synthesize Codex, Gemini, and GPT Pro findings in Chinese.
• Preserve the Plan-only Boundary: score readiness only; do not let the GPT Pro score become an execution trigger.
• Codex remains final owner.
• Do not automate ChatGPT web login.
• Do not read ChatGPT web DOM.
• Do not extract ChatGPT Output programmatically.

Plan-only Boundary

• /ccg:gptpro-plan is planning-only.
• Do not execute implementation.
• Do not apply code changes except writing or updating CCG plan artifacts and GPT Pro bridge artifacts.
• Do not run implementation tasks, mutate product code, commit, push, create a pull request, or continue into /ccg:execute behavior.
• After the user saves GPT Pro output, synthesize Codex, Gemini, and GPT Pro planning findings only.
• Produce or revise the plan, report the plan location and key decisions in Chinese, then stop.
• Stop after producing or updating the plan.
• If the user wants execution, require a separate /ccg:execute <plan> or /ccg:codex-exec <plan> request.

Manual Handoff Barrier

• After creating the bridge artifacts, show only handoff metadata.
• Do not paste the full generated prompt into chat.
• Show the preview URL, session directory, prompt file path, response file path, and status file path.
• Tell the user to open the preview page and use the preview page Copy Prompt button, then manually submit the prompt to ChatGPT Pro and manually save the response.
• End the current assistant turn after the handoff. Do not continue the planning analysis in the same turn.
• Continue only after status.json shows response_saved=true and response.md is non-empty.