audit
/audit
Czyta aplikację Bubble przez oficjalny MCP i pisze plan przeniesienia do kodu z oceną złożoności.
- Autor
- Red-Center-Studio ↗
- Repozytorium
- nocode-migration-audit ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 13.11.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 4
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 13.11.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Czyta aplikację Bubble przez oficjalny MCP i pisze plan przeniesienia do kodu z oceną złożoności.
Argumenty
/audit [app-name]
Oryginalny opis (EN)
Read-only audit of a no-code app before porting it to code. Reads the app's structure (data types, option sets, privacy rules, pages, workflows, plugins) through the platform's official MCP and writes app-spec.json plus a plain-language audit.md with a complexity score, what maps cleanly, what needs work and what does not move. Use when the user asks to audit, assess, inventory or plan porting a no-code app (currently Bubble apps) to their own code or to Cloudflare.
Jak działa
- Rules (apply for the whole run)
- Steps
- Spec checklist
Podgląd SKILL.md
---
name: audit
description: Read-only audit of a no-code app before porting it to code. Reads the app's structure (data types, option sets, privacy rules, pages, workflows, plugins) through the platform's official MCP and writes app-spec.json plus a plain-language audit.md with a complexity score, what maps cleanly, what needs work and what does not move. Use when the user asks to audit, assess, inventory or plan porting a no-code app (currently Bubble apps) to their own code or to Cloudflare.
argument-hint: "[app-name]"
---
# No-code app audit
You are auditing one app the user owns, on their machine, with their own platform login. You read its structure and write two files. You never change the app and you never read its records.
Currently supported source: Bubble®, through Bubble's official MCP server. Other sources will be added as separate readers.
## Rules (apply for the whole run)
1. **Read only.** Use only these five MCP tools: `list_apps`, `get_app_outline`, `get_app_plan_context`, `list_types`, `get_node`. Never call any tool that creates, updates, deletes, applies, installs, uninstalls, uploads, restores, undoes, renames, moves, sets, assigns, detaches, initializes, branches or saves. If you are unsure whether a tool writes, do not call it.
2. **No records.** Never call `search_things`, `get_thing`, `get_app_data_context`, `query_logs`, `query_app_metrics` or any tool that returns user data, logs or metrics. The audit is about structure, not data.
3. **One app.** Audit only the app the user named. If they did not name one, call `list_apps`, show them the names, and ask which one. Never open any other app.
4. **Test version.** Read the development (`test`) version unless the user asks for another version by name.
5. **No secrets.** Do not copy API keys, tokens, passwords or full URLs with query strings into the output. For API calls record the host only. Workflow actions can contain values typed in as plain text (for example a demo account's email and password in a log-in action). Never copy those: write `[literal value omitted]` in their place and list the action under "Needs care" in the audit.
6. **No invented numbers.** Do not give hours, costs or savings. The complexity score is a comparison score, calculated exactly as `reference/audit-template.md` defines it.
7. **Say what you could not read.** Anything you could not resolve goes in `unknowns`, not into a guess.
8. If the MCP is not connected or a read fails, stop and tell the user what is missing. Do not try another route.
## Steps
1. **Pick the app.** Call `list_apps` once. If `$ARGUMENTS` names an app, find it there and take its canonical `appname` and its `displayName` (used as the report title). If no app was named, show the names and ask (rule 3).
2. **Read the app.** Follow `readers/bubble.md` step by step. Keep a running count of every MCP call you make by tool name. Every call must be on the allowed list in rule 1. Reading takes roughly one call per data type, page, reusable, workflow and data-changing action, so the count grows with the app. In testing, an app with 6 data types, 11 pages and 61 workflows took 184 reads. After the outline, tell the user how many pages, reusables, backend workflows and data types the app has, and that the read will take a while; carry on unless they say stop.
3. **Write `app-spec.json`.** It must match `spec/app-spec.schema.json` (spec version `0.1`). Set `source.reader` to `{"name": "nocode-migration-audit", "version": "0.1.1"}`, `source.appVersion` to the version you read (`test` unless the user chose another), and `source.writeCallsMade` to `0`. Before saving, check it yourself against the "Spec checklist" below.
4. **Write `audit.md`.** Follow `reference/audit-template.md` exactly, including the score table.
5. **Save both files** to `migration-audit/<app-name>/` in the current working directory (create the folder). Do not overwrite an existing audit without asking; if one exists, ask whether to replace it or write to a dated folder.
6. **Report back** in a few plain sentences: where the files are, the complexity band, the top three things that need work, the number of unknowns, and the number of MCP calls made (all reads).
If this skill was installed from a clone of the repo and Node.js 22.18 or later is available, the user can run `npm install` then `npm run validate -- migration-audit/<app-name>/app-spec.json` in the repo for a full schema check. Offer it; do not run `npm install` without asking.
## Spec checklist
- Every `key` is snake_case and unique within its list. Field keys are unique within their entity.
- Every `reference` field has `ref` pointing at an entity key in the spec; every `enum` field has `enum` pointing at an enum key.
- Every permission rule names an existing entity, and every field it lists exists on that entity.
- Every page's `workflows`, `usesReusables`, `dataShown` and guard `to` point at keys that exist.
- Workflow steps are numbered 1, 2, 3 in order.
- Conditions and values keep both `text` (readable) and `source` (the expression JSON) whenever the platform returned JSON. A static value with no expression gets the value itself as `text` and no `source`.
- Every field named in a step's `sets`, a rule's `searchableFields` or `autoSaveFields` exists on the right entity.
- No property outside the schema. No secrets, no record data, no personal data.
Pliki w paczce (4)
- SKILL.md 5 KB
- readers/bubble.md 16 KB
- reference/audit-template.md 5 KB
- spec/app-spec.schema.json 15 KB
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · Red-Center-Studio/nocode-migration-audit | 100 | MIT |
| marcel-bich/marcel-bich-claude-marketplace | 100 | MIT |
| jeremylongshore/tons-of-skills-marketplace | 100 | MIT |
| friendlyinternet/nuxt-crouton | 100 | MIT |
| userKael/rastro | 100 | MIT |
| Archont561/qgis-agent-harness | 100 | MIT |
| lilly-logic-labs/lll-skills | 100 | MIT |
| Come-In-Labs/agent-skills | 100 | MIT |
| minanyang/10x-agent | 100 | MIT |
Migawka bazy z 13.11.2026 · wersja paczki 470e6c6e7269 (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100bubble-mcp-field-guide
Prowadzi pracę nad aplikacją Bubble.io przez serwer MCP: pola, strony, reguły prywatności, API i audyt zmian. - 100no-code-app
Pomaga zbudować aplikację na platformie no-code lub low-code (Bubble, Softr, Glide z Airtable lub Xano): wybór platformy, model danych, przewidywalny koszt i plan wyjścia przed lock-in. - 90port-github-app-to-origin
Planuje przeniesienie aplikacji GitHub do Cursor Origin App: odczytuje z kodu jej potrzeby, mapuje je na API Origin i przygotowuje brief. - 100app-audit
Ocenia jedną samodzielnie hostowaną aplikację webową względem wzorca i wystawia raport z priorytetami P0 do P3, tylko do odczytu. - 100store-review-audit
Sprawdza wydanie iOS lub Androida przed złożeniem względem aktualnych zasad sklepów i pisze raport z poprawkami. - 100sandbox-migrate-to-next
Przenosi aplikację Cloudflare Sandbox ze stabilnego pakietu na wersję @next, czyli Sandbox SDK 1.0 w podglądzie.