plan mode: add TL;DR checkpoint and client behavior note (#10195)

## Summary
- Tightens Plan Mode to encourage exploration-first behavior and more
back-and-forth alignment.
- Adds a required TL;DR checkpoint before drafting the full plan.
- Clarifies client behavior that can cause premature “Implement this
plan?” prompts.

## What changed
- Require at least one targeted non-mutating exploration pass before the
first user question.
- Insert a TL;DR checkpoint between Phase 2 (intent) and Phase 3
(implementation).
- TL;DR checkpoint guidance:
  - Label: “Proposed Plan (TL;DR)”
  - Format: 3–5 bullets using `- `
  - Options: exactly one option, “Approve”
- `isOther: true`, with explicit guidance that “None of the above” is
the edit path in the current UI.
- Require the final plan to include a TL;DR consistent with the approved
checkpoint.

## Why
- In Plan Mode, any normal assistant message at turn completion is
treated as plan content by the client. This can trigger premature
“Implement this plan?” prompts.
- The TL;DR checkpoint aligns on direction before Codex drafts a long,
decision-complete plan.

## Testing
- Manual: built the local CLI and verified the flow now explores first,
presents a TL;DR checkpoint, and only drafts the full plan after
approval.

---------

Co-authored-by: Nick Baumann <@openai.com>
This commit is contained in:
baumann-oai
2026-01-30 10:14:46 -08:00
committed by GitHub
parent 5662eb8b75
commit 1ce722ed2e

View File

@@ -37,6 +37,8 @@ When in doubt: if the action would reasonably be described as "doing the work" r
Begin by grounding yourself in the actual environment. Eliminate unknowns in the prompt by discovering facts, not by asking the user. Resolve all questions that can be answered through exploration or inspection. Identify missing or ambiguous details only if they cannot be derived from the environment. Silent exploration between turns is allowed and encouraged.
Before asking the user any question, perform at least one targeted non-mutating exploration pass (for example: search relevant files, inspect likely entrypoints/configs, confirm current implementation shape), unless no local environment/repo is available.
Do not ask questions that can be answered from the repo or system (for example, "where is this struct?" or "which UI component should we use?" when exploration can make it clear). Only ask once you have exhausted reasonable non-mutating exploration.
## PHASE 2 — Intent chat (what they actually want)
@@ -97,6 +99,7 @@ Only output the final plan when it is decision complete and leaves no decisions
The final plan must be plan-only and include:
* A clear title
* A TL;DR section (35 bullets)
* Exact file paths to change
* Exact structures or shapes to introduce or modify
* Exact function, method, type, and variable names and signatures