[ctawiah]

Context

ldcli sync prompts keeps local prompt files in step with the variations of LaunchDarkly AI Configs. Before it writes, it shows a plan: one row for each variation, with the action that sync will take, such as a push or a pull.

The plain-text plan is hard to read today:

• Each row prints project, AI Config, and variation headings, an Action: line, and a full diff. Rows that are in sync print too.
• add --dry-run prints every file in full.

This PR gives the plan, the results, and the add preview a compact layout. Later PRs in this stack add tool and skill rows, and they print them with this layout.

What changes

• The plain-text and Markdown output of the plan, the results, and the add preview use a compact layout. JSON output does not change.
• A summary line comes first. Conflicts and errors get their own clause. A plan with nothing to do prints No changes. N in sync.
• Each change gets one row, grouped under Skills, Tools, or Prompts in apply order. Rows that are in sync appear only in the count.
• The size column counts the added and removed lines in the diff of the row. Placeholder lines such as (does not exist in LaunchDarkly) do not count.
• On a color terminal, the size column shows +N in green and -N in red. These are the colors of the added and removed lines in a diff.
• Piped output, -o markdown, -o json, and output with NO_COLOR set have no color. Column widths ignore the color codes, so the columns stay aligned.
• A new --diff flag prints each full diff. A conflict always shows its diff. sync prompts, watch, add, attach, and link have the flag.
• A run that is not a dry run prints the plan only when it needs it. That is before the confirmation question, when a conflict or error row stops the run, or when there is nothing to apply. With --yes, only the results print. A dry run always prints the plan.
• A conflict prints one header line, such as Conflict: variation demo/support/concise changed on both sides., and then its diff.
• The note about another working copy (another checkout of the repository) is one Note: line on stderr.
• -o markdown prints one table for each kind. add --dry-run lists the files and their line counts.
• BlockingError names the kind of the row, for example cannot sync conflicted resource variation demo/support/concise. Existing error substrings still match.

A plan now looks like this:

2 changes: 1 prompt push, 1 prompt pull. 1 in sync.

Prompts
  pull   demo/support/concise   instructions changed in LaunchDarkly   +1 -1
  push   demo/support/default   instructions changed                   +1 -1

And the results of ldcli sync prompts --yes:

Prompts
  pulled   demo/support/concise   instructions changed in LaunchDarkly
  pushed   demo/support/default   instructions changed

2 applied. 1 in sync.

Design notes

• One routine, diffBlocks in render_diff.go, feeds both the diff renderer and the size counter. So a size always matches what --diff shows.
• Placeholder lines stay out of the size. If they counted, a create reads +5 -1 although it removes nothing.
• The words and reasons come from small tables and functions in output.go. writeOutcomeOutput takes the executed plan, so a result row can show the plan reason.
• This PR removes the unused field conflictGroup.attachment. It always equals "more than one resource".
• Result rows do not show the new pins of a variation yet, such as pins search v4. A pin is the tool or skill version that a variation uses. That data comes from the apply step, so feat: apply tool and skill rows in a sync #941 adds it.

Review focus

• Does the plan still print in every case where the user needs it?
• Does each size match the --diff output of its row?
• Is the JSON output of the plan and the results unchanged?
• Do scripts that match BlockingError text still match?

Verification

• go test ./...

• golangci-lint run (v2.9.0)

• TestPlanAndOutcomeJSONIsStable compares the plan and outcome JSON with a saved copy, so a later change cannot change the JSON by accident.

• TestWritePlanReviewDiffMatchesRowSize compares each size with the --diff output.

• I built ldcli and ran it against a local stand-in for the LaunchDarkly API. The runs covered add --dry-run with and without --diff, add, a dry run, --yes, and -o markdown. They also covered a conflict with and without --conflict local, an error row, and a stale working copy.

• TestDiffSizeTextColorsEachPart, TestWritePlanReviewColorsSizesOnATerminal, and TestWritePlanSizesHaveNoColorOffATerminal cover colored and plain sizes. The second test removes the color codes and compares the result with the plain plan.

• I ran a dry run in a 120-column pseudo-terminal. +1 was green and -1 was red. With NO_COLOR=1, with -o markdown, and through a pipe, the plan had no color.

Related changes

1. This PR: show a compact sync plan and add a --diff flag
2. feat: add target flags to sync prompts and read sync flags only from the command line #938: add target flags, and read sync flags only from the command line
3. feat: plan standalone tool and skill rows by sync scope #939: plan tool and skill rows by sync scope
4. feat: create skills during sync, and read again after a 409 on create #940: create skills during sync, and read again after a 409 on create
5. feat: apply tool and skill rows in a sync #941: apply tool and skill rows in a sync

---

Note

Overview
Reworks plain-text and Markdown output for ldcli sync prompts so plans and apply results are scannable: a one-line summary, then aligned rows grouped under Tools / Prompts (action, resource, reason, optional Size +N -N from the same diff logic as rendering). In-sync resources only appear in the count.

Adds --diff on sync, watch, add, and apply-style commands to print full diffs (conflicts always show a diff). add --dry-run lists paths with line counts; with --diff, file bodies follow. Plan printing is lazy: shown before confirm, on blocking errors, when nothing applies, or on dry-run; --yes skips the plan and prints compact results only. Stale-working-copy hints move to stderr Note: lines; BlockingError aggregates all blockers and uses ResourceID.Label().

JSON plan/outcome shape is unchanged (covered by stability tests). Terminal size colors respect NO_COLOR and non-TTY output.

^{Reviewed by Cursor Bugbot for commit a576171. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit fb46607. Configure here.}