Show each plan step's acceptance checks on the review card - #50
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why this was done
The plan review card is the one moment where the user decides whether to let MandoCode run a multi-step job. Until now that card showed only what each step would do — not how anyone would know it worked.
The CLI side of this change (see the linked MandoCode PR) makes acceptance criteria a first-class part of every plan step: 2–5 concrete, observable checks that define "done" for that step, and that the executor is held to. Those criteria were being generated and enforced, but the Desktop user never saw them. That is the wrong place to hide them — the approval screen is exactly where they matter.
This PR surfaces them, and fixes a related correctness gap in the plan editor.
What changed
1. The plan review card now shows each step's acceptance checks.
Under each step's instruction, the card renders a numbered "Acceptance checks" list. The user can see, before approving, what the agent will actually be measured against — and can reject a plan whose checks are too weak or off-target, rather than discovering that after the run.
2. Editing a step's instruction now resets its acceptance checks.
Previously, editing a step's instruction left the old acceptance criteria attached to it. That produced a genuinely misleading state: the card would show the user's new instruction paired with checks written for the instruction it replaced, and the executor would be graded against the stale ones. Now an edited step's criteria are reset to the instruction the user actually typed, and its "evidence follow-up already used" flag is cleared so the edited step gets a clean run rather than inheriting the prior attempt's budget.
3. Submodule pointer bump to pick up the CLI-side planner work.
Impact for users
Risk
Low. The rendering change is additive and skips cleanly when a step has no criteria. The editor change touches only the step the user just edited.
Merge order
This PR bumps the
MandoCodesubmodule to the branch commit from the linked CLI PR. Merge the MandoCode PR first, then update this branch's submodule pointer to the resulting commit onmainbefore merging here — otherwise Desktopmainwould point at a commit that isn't on CLImain.Related
Depends on the MandoCode CLI PR that introduces acceptance criteria and the final test-and-repair phase.