[PF-24] Run plan and implement in foreground with streaming in Cursor sandbox

[chucoding]

Purpose

Run both /p (plan) and /i (implement) in the foreground in the Cursor chat sandbox so that their output streams in the agent terminal, instead of running them in the background.

Description

• p skill (.cursor/skills/p/SKILL.md, templates/cursor/skills/p/SKILL.md): Plan script is run in the foreground so output streams in the Cursor chat sandbox terminal. Step 3 restored to "After it completes" — read the generated .plan.md and summarize in the reply.
• i skill (.cursor/skills/i/SKILL.md, templates/cursor/skills/i/SKILL.md): Implement script is run in the foreground (no longer in the background). Step 3 changed to "After it completes" — on success summarize or suggest next steps; on failure report the error.
• workflow rule (.cursor/rules/workflow.mdc, templates/cursor/rules/workflow.mdc): For /p, added explicit instruction to run the command in the foreground so output streams in the agent terminal (sandbox).

Location	Change
p skill step 2	"in the background" → "in the foreground"; "output streams in the Cursor chat sandbox terminal"
p skill step 3	"After starting the command" / timeout guidance → "After it completes" + read .plan.md and summarize
i skill step 2	"in the background" → "in the foreground"; removed "Do not wait for completion", short timeout note
i skill step 3	"Immediately after starting it" / end turn → "After it completes" + summarize or suggest next steps

Notion: The Cursor Sandbox page was updated with a new section "6. Cursor 스킬 실행 방식" describing that plan and implement run in the foreground with streaming (no code change in this repo for that).

How to test

1. In Cursor Agent chat, run /p with a short goal (e.g. /p add a readme).
2. Confirm the plan command runs in the foreground in the sandbox terminal and that plan output streams there.
3. After completion, confirm the agent reads the generated .plan.md and summarizes it in the reply.
4. Run /i with a short prompt and confirm implement runs in the foreground with streaming; after completion, confirm the agent summarizes the result or suggests next steps.

Review Requirement

• Confirm p and i skill wording is consistent (foreground, streaming, "after it completes").
• Confirm template copies under templates/cursor/ match the .cursor/ versions.

Additional Info

• Related Notion: Cursor Sandbox — section "6. Cursor 스킬 실행 방식" added for this behavior.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.