agentleFS
DevinoSolutionsDevinoSolutions/claude-plugins/skills/reviewing-visual-builds/SKILL.md

reviewing-visual-builds

Skill0 starsChanged 9 days agoMITView on GitHub

Reviews a SnapVisor visual regression build: shows what changed in the screenshot diffs for a branch or commit and approves or rejects the build after the user confirms. Use when the user asks 'what changed in the latest build', 'show the screenshot diffs', 'review the visual changes on this branch', 'approve build 42, the nav restyle is intentional', 'reject this build', or asks about visual testing or UI regression results.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 3 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Operations listed by name: call them directly, for example snapvisor:listBuilds.

    line 16
  2. Very specific:

    If the connection lists the operations by name instead (the per-tool surface), call them directly without the external_ prefix.

    line 27
  3. Very specific:

    If you do not know the account and project, run one program: external_getMe({}), then external_listProjects for the account the user means.

    line 41
  4. Distinctive:

    If the user wants a decision, restate it ("Approve build 42 in project web.

    line 45
  5. Very specific:

    Call external_createReview with event: "APPROVE" or event: "REJECT" and the user's reason as body.

    line 46
  6. Distinctive:

    Never file a review without the user's explicit confirmation in this conversation, and never approve or reject several builds in a loop.

    line 63
  7. Distinctive:

    Always show the diffs before asking for a decision.

    line 64

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md
<!-- Generated from plugins/snapvisor/skills/reviewing-visual-builds/SKILL.md by scripts/sync-skills.mjs. Edit the plugin copy, then run the script. -->

Review a SnapVisor build

Walk the user through the visual changes SnapVisor detected in a build, then approve or reject it on their say-so. An approval is a real decision for the whole team: the build is accepted and its changed screenshots become the new baseline.

Connecting and calling the tools

Every operation in this skill comes from the snapvisor MCP server, the SnapVisor connector. The server defaults to Code Mode, so the operations below are written as the external_* functions you call inside snapvisor:execute_typescript. On a server set to the per-tool surface, the same operation is listed as snapvisor:<name> without the external_ prefix, for example snapvisor:listBuilds.

  • Only snapvisor:search_tools and snapvisor:execute_typescript listed (Code Mode, the default): call snapvisor:search_tools first, then call each tool inside snapvisor:execute_typescript as external_<tool> (for example external_listBuilds); the program must return its result. Scopes, confirmations and gates are identical on both surfaces.
  • Operations listed by name: call them directly, for example snapvisor:listBuilds.
  • "Tool not found": switch to the Code Mode route instead of retrying the direct call. If snapvisor:search_tools does not return the tool either, its OAuth scope was not granted; the user reconnects SnapVisor and approves that scope.
  • No snapvisor tools at all: SnapVisor is not connected yet. Follow the setup-snapvisor skill.

Tools you will use

The snapvisor MCP server has two tools. Everything below goes through them.

  • snapvisor:search_tools: find the operations you need and their exact input shape. Call it before your first snapvisor:execute_typescript, for example with the queries build, diff, review.
  • snapvisor:execute_typescript: run a short program that calls the external_* functions snapvisor:search_tools declared. They are already in scope; never import or redeclare them. The program must return its result.

If the connection lists the operations by name instead (the per-tool surface), call them directly without the external_ prefix.

Operations this skill uses (confirm names and inputs with snapvisor:search_tools):

  • external_getMe: the signed-in user and the accounts (with slug) this connection can reach.
  • external_listProjects: projects in an account.
  • external_listBuilds: builds of a project, newest first; filter by branch (head), commit (headSha) or search.
  • external_getBuild: one build with status, conclusion, stats, branch and commit.
  • external_listBuildDiffs: one entry per screenshot with name, status (added, changed, removed, failure, ignored and more), and base and head image links. Pass needsReview to see only what needs a decision.
  • external_listReviews: reviews already filed on the build.
  • external_createReview: file a review with event APPROVE or REJECT and an optional body.

Workflow

  1. If you do not know the account and project, run one program: external_getMe({}), then external_listProjects for the account the user means. Ask once if more than one project fits.
  2. Find the build. With no build number given, take the newest from external_listBuilds (or the newest on the branch the user named).
  3. In one snapvisor:execute_typescript call, fetch external_getBuild, external_listBuildDiffs and external_listReviews together with await Promise.all([...]) and return a compact summary.
  4. Present the build number, branch, commit and status, then the changed screenshots grouped by status (changed, added, removed, failed). Give each its before and after links. Mention reviews already on the build.
  5. If the user wants a decision, restate it ("Approve build 42 in project web; 6 changed screenshots become the baseline") and wait for a clear yes.
  6. Call external_createReview with event: "APPROVE" or event: "REJECT" and the user's reason as body. Report the new review, and confirm the build's status with external_getBuild.

Code Mode pattern

const [build, diffs, reviews] = await Promise.all([
  external_getBuild({ owner, project, buildNumber }),
  external_listBuildDiffs({ owner, project, buildNumber, needsReview: "true" }),
  external_listReviews({ owner, project, buildNumber }),
]);
return { build, diffs: diffs.results.map((d) => ({ name: d.name, status: d.status })), reviews };

Wrap a call in try { ... } catch (e) { return String(e) } when it may fail. Errors read <CODE>: <details>; relay the code and details instead of guessing.

Rules

  • Never file a review without the user's explicit confirmation in this conversation, and never approve or reject several builds in a loop. One build, one confirmation.
  • Always show the diffs before asking for a decision.
  • Image links are signed and expire after about an hour. Hand them over; do not store or archive them.
  • A 403 saying MCP access requires a paid plan means the account is on the Free plan. Say so plainly and do not retry.
  • SnapVisor compares screenshots the user's CI uploaded. Do not offer to capture, upload or edit a screenshot; no operation on this surface can.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.