SkillFoxx

Type a name, a task or a service

Match with Foxx AI

1c-testpilot-agent-kit skill

A skill and tester agent for Claude Code that test the 1C interface through the 1c-testpilot MCP server with evidence

  • 22 ★ on the repository
  • Last commit
Skill

Extended access

Why this level

  • The agent drives the 1C test client: it fills forms, posts documents and changes database data.
  • A run on a working database leaves real documents behind, so test on a copy.
All reasons and checks
Russian stack

afk-yar/1c-testpilot-agent-kitAuthor:afk-yar

Install

In your terminal, with SkillFoxx CLI

npx skillfoxx add skills/1c-testpilot-agent-kit

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • The agent drives the 1C test client: it fills forms, posts documents and changes database data.
  • A run on a working database leaves real documents behind, so test on a copy.

This is third-party code. Review the repository files before installing.

What it does

The kit builds on the 1c-testpilot MCP server, which connects to the 1C test client over its network protocol. The `testpilot-1c` skill describes the canonical workflow: opening a form, filling attributes and table parts, pressing a button and reading the error window. It gathers 11 pitfalls found in runs, an observation mode for the session, a Python API with pytest for repeatable checks and context-saving techniques. A separate `agents/tester-1c.md` file defines the tester agent: its role, a verdict with evidence and work through a session folder. The repository also holds a benchmark on 1C:ERP with an order, invoice and payment scenario and scripts that count tokens and cost from Claude Code and Codex CLI logs.

Who it is for. QA engineers and 1C developers who test the 1C interface with agents and have installed or are ready to install 1c-testpilot

Good fit when

  • An order, invoice and payment scenario needs an agent run and a verdict with evidence.
  • An agent spends too many tokens on the form tree and needs snapshots with state differences.
  • A scenario the agent has passed should become a pytest test that runs without a model.
  • The 1C test client hits known pitfalls and their descriptions should be at hand.

Not a fit when

  • The 1c-testpilot server is not installed yet: the kit does not work without it.
  • You are testing BSL code and leaving the 1C interface alone: other tools cover code.
  • You need a ready test suite for your configuration: the benchmark targets 1C:ERP 2.5 with an English interface.

Example request

Run the scenario: create a customer order, issue an invoice and a payment, then verify the documents with a database query

Limitations

The kit does not contain 1c-testpilot itself; it is installed separately per the author's README and is distributed under AGPL-3.0. The skill and method are written in Russian. The pitfalls were verified on testpilot 1.7.1 and platform 8.3.27, and some may go stale in newer versions. The benchmark ran on the English interface of 1C:ERP 2.5.27 with one run per variant, so it is an observation and not a model ranking. The tester agent is a pilot with about ten runs behind it.

How to disable. Delete the `testpilot-1c` directory from the skills directory and the `tester-1c.md` file from `~/.claude/agents/`.

FAQ

Do I need to install 1c-testpilot separately?

Yes. The kit only adds a skill and an agent on top; the ROCTUP MCP server is installed per its README and then connected to the agent.

Does the kit work with Codex?

Yes, give the agent the path to `SKILL.md` in the prompt and it reads it itself. Usage-counting scripts also exist for Codex CLI logs.

What is inside the benchmark?

A hard scenario with six traps on 1C:ERP 2.5, a 16-point scoring method and protocols of eight Claude and Codex runs.

What does the tester agent do?

The agent from `agents/tester-1c.md` runs a scenario, returns a verdict with evidence and works through a session folder. It is a pilot with about ten runs.

What do the benchmark results show?

On one scenario the agent with testpilot used about 62% of the tokens of the agent with Tester and finished in half the time. That is one run per variant.

Discussion

No questions or reviews yet.

All 1C:Enterprise tools

Tester

Тестер

Russian stack

A scenario testing system for 1C:Enterprise with a set of agent skills that let an AI agent operate the application like a regular user

SkillBroad access207
Russian stack

An MCP server that drives the 1C test client's interface directly: opens forms, fills fields, records and replays scenarios

MCP serverBroad access80
Russian stack

A 1C development skill collection for Codex and Claude Code: XML sources, forms, data composition, infobases, web publishing and an HTTP test bridge

SkillBroad access62
Russian stack

A Claude Code skill for driving the 1C web client through Playwright MCP: sessions, navigation, catalogs, documents and test data

SkillBroad access33
Russian stack

A set of 52 Claude Code skills and agents, plus Cursor rules, for the full 1C:Enterprise 8.3 development cycle, assembled from several community projects

SkillExtended access

README in short

The README covers 1c-testpilot's strengths, lists the kit's four folders and gives short benchmark results. It then offers a three-step quick start and a list of limitations.

README badge

Original SKILL.md

---
name: testpilot-1c
description: UI-тестирование 1С через MCP-сервер 1c-testpilot (ROCTUP). Используй когда нужно управлять тестовым клиентом 1С напрямую — открывать формы, заполнять реквизиты, работать с табличными частями, сравнивать состояние формы до и после.
argument-hint: "сценарий на естественном языке"
---

# /testpilot-1c — UI-тестирование через 1c-testpilot

Сервер ходит напрямую в `/TESTCLIENT` по протоколу тестового клиента. Базы-менеджера нет,
BSL-сценарии писать не надо, один шаг — один вызов инструмента.

Своего скилла у автора нет, вся справка лежит в описаниях инструментов и в `docs/TOOLS.md`.
Этот файл — только то, чего в описаниях нет или что там сформулировано неоднозначно.

## Окружение

Все пути и настройки конкретной машины собраны здесь; ниже по тексту — только имена.
Заполните таблицу своими значениями.

| Что | Где |
|---|---|
| Сервер testpilot (клон ROCTUP/1c-testpilot) | `<каталог testpilot>` — проверено на 1.7.1 |
| Точные сигнатуры действий | `docs/TOOLS.md` сервера |
| Python API и pytest | `docs/PYTHON_TESTING.md` сервера |
| Типы значений для обработки Testpilot | `onec/Testpilot/README.md` сервера, раздел «Типы значений» |
| Профили запуска | `profiles.yaml` сервера (образец — `profiles.example.yaml`) |
| Порты | свой порт на каждое назначение: агенты, скрипты, pytest |
| Образцы повторяемых прогонов | ваши скрипты и тесты на Python API |
| Снятые грабли 5, 6, 10, 12, 13 | `история-граблей.md` рядом с этим файлом |

## Инструменты

Схемы отложенные, тянуть через `ToolSearch` запросом
`select:mcp__1c-testpilot__tc_session,...`. Все десять групп разом — около 21 тыс.
токенов, поэтому грузи только нужные. Базовый набор на обычный сценарий:
`tc_session`, `tc_find`, `tc_window`, `tc_field`, `tc_table`.

## Канонический поток

1. `tc_session` action=launch_client, profile=`<проект>` — профили в `profiles.yaml` сервера.
2. `tc_find` action=wait_for_object_displayed, cls=`ManagedForm` — дождаться интерфейса.
3. `tc_window` action=execute_command, command=`e1cib/list/<Тип>.<Имя>` — открыть список.
4. `tc_find` action=find_object / find_objects — получить ref нужных элементов.
5. `tc_field` / `tc_table` — действия.
6. `tc_window` action=close_window (можно с `ref` окна). Если на форме вопрос о сохранении,

Report a problem

Foxx AI1c-testpilot-agent-kit
Foxx AI

I am Foxx AI and I have already vetted this tool. Ask about install, setup or anything else, and I will keep it simple.