commerce-app-review — Claude Code skill: review and install
commerce-app-review reviews an App Builder project against Adobe Commerce submission guidance and separates blocking findings from recommendations. Use it before preparing an Exchange submission.
Apache-2.0 205 installs ★ 335 updated 2026-10-24
npx skills add adobe/skills --skill commerce-app-review -a claude-codeWhat commerce-app-review does
This skill checks a local App Builder project against Adobe’s Commerce submission guidance. It reads project configuration, dependencies, environment examples, documentation and UI code, then ties reported findings to files and line numbers. Checks that require npm audit, a live Exchange listing or a deployed environment are excluded.
Provide an app type, either downloadable or non-downloadable. The project path is optional; when omitted, the current working directory is used. The skill fetches Adobe guidance with curl and reads local reference files when present. If aio is missing, the review continues, though walkthrough details use fallback knowledge. A failed guidance page also limits coverage, so the skill warns when that happens.
It first prints MUST blockers and NICE recommendations, then waits for you to choose a walkthrough or a finding code. A finding needs a specific file and line before it can be reported. You can then discuss how to address it.
What it helps with
- check an App Builder project before submitting it to Adobe Exchange
- spot security issues in application code
- compare project structure and documentation with Adobe guidance
- trace dependency findings to the relevant files
Example prompts for the agent
- Review this App Builder project before an Adobe Exchange submission. App type:
downloadable. - Check the project in the current directory. The app type is
non-downloadable. - Walk me through finding
MUST-2and help me address it.
How to install commerce-app-review
1Skills CLI (Vercel)
npx skills add adobe/skills --skill commerce-app-review -a claude-code2Manually: put the files in place
~/.claude/skills/commerce-app-review/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/commerce-app-review/
1Skills CLI (Vercel)
npx skills add adobe/skills --skill commerce-app-review -a codex2Manually: put the files in place
.agents/skills/commerce-app-review/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add adobe/skills --skill commerce-app-review -a cursor2Manually: put the files in place
.cursor/skills/commerce-app-review/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install adobe/skills plugins/commerce/app-review/skills/commerce-app-review2Manually: put the files in place
.github/skills/commerce-app-review/SKILL.md1Skills CLI (Vercel)
npx skills add adobe/skills --skill commerce-app-review -a gemini-cli2Platform CLI
gemini skills install https://github.com/adobe/skills --path plugins/commerce/app-review/skills/commerce-app-reviewManually: put the files in place
.devin/skills/commerce-app-review/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add adobe/skills --skill commerce-app-review -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about commerce-app-review
How do I install commerce-app-review in Claude Code?
Run npx skills add adobe/skills --skill commerce-app-review -a claude-code. The skill file is placed at ~/.claude/skills/commerce-app-review/SKILL.md.
What does commerce-app-review do?
It checks a local App Builder project against Adobe Commerce submission guidance and sorts results into MUST blockers and NICE recommendations. Findings include a file path and line number.
What do I need to provide before a review?
Choose downloadable or non-downloadable for the app type. A project path is optional; otherwise, the current working directory is used.
Source text
SKILL.md · Apache-2.0
---
name: commerce-app-review
description: Use when validating an App Builder app before submitting to Adobe Exchange, or when the user mentions Adobe Commerce app compliance, extension review, marketplace submission, or Adobe Exchange listing. Checks security requirements, project structure, documentation completeness, and dependency health; reports MUST-have blockers and NICE-to-have recommendations and walks through each finding interactively.
license: Apache-2.0
---
# Self-review an App Builder app
Reviews a local app directory against Adobe's Commerce submission guidelines and walks through each finding interactively — MUST-have blockers first, then NICE-to-have recommendations.
## Inputs
- `APP_PATH` *(optional)* — path to the app root. Defaults to the current working directory.
- `APP_TYPE` *(required)* — `downloadable` or `non-downloadable`. A downloadable app allows merchants to download the source package; a non-downloadable app is installed directly from Adobe Exchange with one click.
## Procedure
### 1. Validate the path and app type
If `APP_PATH` is not provided, use the current working directory. Confirm it exists and is a directory. If not, stop and report:
> Path not found. Please provide the path to the root of your app codebase.
If `APP_TYPE` is not provided, ask:
> **Is this app downloadable or non-downloadable?**
> - **Downloadable** — merchants download the source package and deploy it themselves
> - **Non-downloadable** — merchants install directly from Adobe Exchange with one click
Do not proceed until both are confirmed.
Check whether `aio` is installed (e.g. `command -v aio`). This is a presence check only — do not attempt to authenticate at this stage. If `aio` is not found, note it once:
> Note: `aio` isn't installed. Walkthrough enrichment (Step 8) will use fallback knowledge
> unless `aio` is installed and authenticated before then.
Continue either way.
### 2. Load local references
Read `references/finding.md` and all files under `references/exceptions/` (paths relative to this skill file).
Load pattern files conditionally:
- references/patterns/cors-storefront-actions.md — always
- references/patterns/accs-rest-api-dependency.md — always
- references/patterns/aio-commerce-sdk.md — always
- references/patterns/admin-ui-sdk.md — only if commerce/backend-ui/1 is in app.config.yaml or ext.config.yaml, or @adobe/uix-guest/@adobe/uix-core in package.json
Pattern files take precedence over fetched guidelines. Exception files suppress findings from both.
If any file is not found, skip and continue.
### 3. Fetch the guidelines
Fetch both pages using `curl`. Use raw GitHub URLs — the rendered developer.adobe.com pages are not reliably accessible:
- **Submission guidelines**: `https://raw.githubusercontent.com/AdobeDocs/commerce-extensibility/refs/heads/main/src/pages/app-development/app-submission-guidelines.md`
- **Admin UI SDK checklist**: `https://raw.githubusercontent.com/AdobeDocs/commerce-extensibility/refs/heads/main/src/pages/admin-ui-sdk/app-review-checklist.md` — only if `commerce/backend-ui/1` is in `app.config.yaml` or `ext.config.yaml`, or `@adobe/uix-guest`/`@adobe/uix-core` in `package.json`
If a page is unreachable, note which one failed, proceed with the other, and warn the user that coverage is partial.
### 4. Read the app
Read these files in order:
1. `app.config.yaml` — extension points, actions, inputs, annotations, productDependencies
2. `package.json` — name, description, author, version, dependencies, scripts
3. `env.dist` / `env.example` / any `env.*` at the root — documented env vars
4. `README.md` — installation instructions, events, API requirements
5. `deploy.yaml` — app ID, APIs, workspaces
6. `install.yaml` — extension points (if Admin UI SDK detected)
7. Every `*.config.yaml` and `ext.config.yaml` under `src/` and `actions/`
8. Every `index.js` under `actions/`
9. `web-src/src/` — UI code, innerHTML, routing
10. `extension-manifest.json` — if Admin UI SDK detected
If a file does not exist, note it and continue.
### 5. Review
Build a complete list of checks from the pattern files and fetched guidelines. Then work through every item silently, one by one. Do not skip any item.
**Exclude the following entirely — do not check:**
- `npm audit`, `npx npm-check`, or any check requiring running npm
- Exchange listing URL, repository visibility, screenshots in Exchange listing
- Any check requiring a live Exchange listing, deployed environment, or external URL
For each item:
- Check it against the files read in Step 4
- If confirmed → record as FINDING with severity, file, and line number
- If an exception in `references/exceptions/` applies → suppress it
- If not applicable to `APP_TYPE` → skip it
**Before recording any FINDING, verify it in the actual files. If you cannot point to a specific file and line number, do not flag it.**
**Before flagging a missing package — verify it is directly imported in source files, not just a transitive dependency.**
After working through all items, collect all FINDINGs and assign codes: `MUST-<n>` and `NICE-<n>`. Merge closely related findings before assigning codes.
For each finding track:
- `SEVERITY` — `MUST` or `NICE`
- `DESCRIPTION` — one sentence, specific and actionable
- `FILE_PATH` — **absolute** path to the relevant file; empty for general issues. Render as a markdown link: `[relative/path](file:///absolute/path)`
- `LINE_NUMBER` — specific line; 0 for general issues
- `CATEGORY` — one or more of: `security` · `documentation` · `code` · `dependencies` · `other`
### 6. Present the raw findings list
Print a concise findings list — **no enrichment yet**. Format:
```
Found N MUST-have issue(s) and M NICE-to-have recommendation(s).
**MUST — blockers**
- MUST-1: <one-line description> (`<FILE_PATH>`, line <LINE_NUMBER>)
- MUST-2: …
**NICE — recommendations**
- NICE-1: <one-line description> (`<FILE_PATH>`, line <LINE_NUMBER>)
- NICE-2: …
```
Omit a group heading if it is empty. Omit the file reference when `FILE_PATH` is empty.
After the list, show:
> Type **walkthrough** to go through each finding · **MUST-n** or **NICE-n** to dig into a specific one · or ask anything.
Then wait for the user's response.
### 7. Respond to the user
- **"walkthrough"** (or equivalent intent) → go to Step 8
- **A finding code** (e.g. "MUST-2", "NICE-1") → enrich that finding and present it; show the actions line again and wait
- **Anything else** → answer it; show the actions line again and wait
Repeat until the user ends the session or all findings have been walked through.
### 8. Walkthrough mode
Work through all findings sequentially — MUST first, then NICE. Before starting, obtain the IMS token (see below).
- If the group is empty, say so and move on.
- For each finding in code order:
1. Show a numbered header (e.g. `--- MUST issue [n/N] ---`)
2. Enrich the finding and print it
3. Show: `**"resolve it"**` to work on this finding together · `**"next"**` to move on · or ask anything
4. Wait for the user's reply
After the last finding, close with a short encouraging message including how many findings were resolved (e.g. "X of N+M findings resolved") and wish the developer good luck.
---
## Obtain IMS token
Run once per session, the first time enrichment is needed:
```bash
IMS_TOKEN=$(aio auth login --bare 2>/dev/null)
```
- If `$IMS_TOKEN` is non-empty → use the documentation agent for enrichment.
- If empty → warn the user; instruct them to run `aio login` and restart for the best experience; ask whether to continue without enrichment or stop.
## Enrich a finding
1. **If IMS token is available**, query the documentation agent:
```bash
curl -s -X POST https://commerce-docs-prod-endpoint-d0ctgyebe7bec8e6.a02.azurefd.net/api/query \
-H "Authorization: Bearer $IMS_TOKEN" \
-H "Content-Type: application/json" \
-d '{"query": "<topic derived from the finding>", "count": 5}'
```
Pick the `index` based on category:
- `security`, `deployment`, `code` → `app-builder-docs`
- `documentation`, `dependencies` → `commerce-extensibility-docs`
- default → omit `index`
2. **Read the actual file** at `FILE_PATH` and `LINE_NUMBER`. Generate a concrete before/after fix specific to the developer's code — not a generic example.
3. **Compose the enriched finding** using the format in `references/finding.md`.
If the agent returns no results, fall back to App Builder / Commerce knowledge and continue.
## Failure modes
- **Path not found** → stop and report; do not proceed
- **Guidelines page unreachable** → note which failed; proceed with the other; warn coverage is partial
- **`aio` not installed or not authenticated** → warn the user, instruct them to run `aio login`, and ask whether to continue without enrichment or stop
- **Documentation agent returns no results** → fall back to App Builder / Commerce knowledge; do not stop
© adobe · Apache-2.0 · adobe/skills