compare
CommandCompare two or more options side by side, recommend one, and let the reader pick in the page
Generate a visual diff review for code changes, with each fix to accept or leave in the page
> /plugin marketplace add nicobailon/visual-explainer> /plugin install visual-explainer@visual-explainer-marketplace
name: diff-review description: Generate a visual diff review for code changes, with each fix to accept or leave in the page
Load the visual-explainer skill and make a visual diff review.
Scope: read `$@` as a branch, commit, range, PR, or `HEAD`. If it is empty, compare the working tree with `main`/`master`.
Gather before you draw: diff stats, name-status, the full changed files and the code paths around them, public API or type changes, tests, dependency or config changes, and commit messages. Every claim cites a path or `file:line`. Do not invent a rationale.
Typical sections; merge, reorder, drop, or add as the content needs:
| Section | Figure | |---|---| | Verdict | one sentence + merge/blocked chip + `<ve-stack>` of lines added and removed | | What changed | file map with added/modified/deleted chips; `<ve-bars>` of lines per file | | Behavior | before/after of the same SVG, edges added or removed highlighted; a changed state machine or queue replays one event per step (`references/motion.md`); a machine read from code is walked before and after, showing a rule the diff newly breaks or a state it makes unreachable (`references/machines.md`) | | Risks | severity-chip table: correctness, tests, API, security, performance | | Coupling | small SVG of what depends on the changed code | | Fixes | each risk worth acting on: `<ve-ask id="r1" q="Fix the unchecked null in parse()?">` (then `r2`, …; the risk in a few words) with `<ve-opt value="fix" default note="the fix in one line">Fix</ve-opt>` and `<ve-opt value="leave">Leave as is</ve-opt>` |
Colors: red = removed or before · green = added or after · amber = risk · blue = context. Deliver with the skill's rules and hand over in one line.
When the reply arrives: on a branch the user owns, apply only the fixes the reader picked; a question they left alone is not a yes. On someone else's PR, draft review comments for them and do not post them.
An agent skill that turns complex terminal output into styled HTML pages you actually want to read. Ask your agent to explain a system architecture, review a diff, or compare requirements against a plan.
Compare two or more options side by side, recommend one, and let the reader pick in the page
Verify a generated document against actual code and git history
Generate a narrated MP4 explainer, or record a walkthrough of a live web app
Generate an interactive visual implementation plan the reader can answer in place
Generate a figure-first HTML page for any topic and open it in the browser
Explain code in detail - what it does, how it works, and why
Create smart commit with quality checks and conventional format
Meta-skill that loads all llmwiki governance docs and exposes the three maintainer slash commands.