RDV AIR CLI
A 1C code review agent: an auditor collects findings, a critic verifies them, and the result is a Markdown report
- 31 ★ on the repository
- Last commit
Extended access
Why this level
- 1C source code goes to the chosen model and leaves the company perimeter when the model is cloud-hosted.
- The CLI reads git history and creates task directories with copies of files before and after changes.
Install
- Works in
- Services
Manual install
Установи и настрой RDV AIR.
Send this to the agent in an open project: the `review-setup` skill installs the `review-tasks` package and sets the Jira prefix.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- 1C source code goes to the chosen model and leaves the company perimeter when the model is cloud-hosted.
- The CLI reads git history and creates task directories with copies of files before and after changes.
This is third-party code. Review the repository files before installing.
What it does
RDV AIR reviews 1C:Enterprise BSL modules and SKD queries by task number or by GitLab Merge Request. The review runs in two stages. In the first an auditor collects candidate findings; in the second a critic verifies each against the code and sources and confirms or rejects it. The model calls commands that write candidates, evidence and decisions into a YAML protocol, and the engine checks required fields and builds a `review-report-<task>.md` report with links to GitLab lines. Task changes come from git history by a task key like `TASK-123` in commit messages. Through MCP the agent checks its output against platform documentation, development standards and BSP, but the MCP servers themselves are connected separately. A review can be started from Codex, Cursor, OpenCode or a button in the Jira extension. Formatting and typos are left to SonarQube and linters.
Who it is for. 1C development teams on GitLab and Jira that want an automatic code review before handing a task to testing
Good fit when
- Before handing a task to testing you want to run a 1C code review yourself and fix the findings.
- An architect checks a Merge Request and needs a report with links to GitLab lines.
- Reviews should start from a Jira issue with one button.
- SKD queries are checked for performance against metadata and indexes.
Not a fit when
- Code cannot go to a cloud model and there is no local model: the agent passes code to the chosen model.
- You need automatic comments in GitLab: for now the result is only a Markdown report.
- You are checking formatting and typos: SonarQube and linters handle that.
Example request
Check the changes for task TASK-123 and prepare a report with findingsLimitations
The agent passes the code under review to the chosen model, cloud or local, so the code policy should be settled first. MCP servers are not included and are connected separately, though they can be turned off. Task changes are found through a local git repository by a key in commit messages, and the GitLab API is needed only by the Jira extension. The agent does not leave GitLab comments yet. The license is stated as BSD 3-Clause, while GitHub detects it as non-standard.
How to disable. Remove the package with `python -m pipx uninstall review-tasks` and delete the task directories created by `init`.
FAQ
How does RDV AIR get changes from GitLab?
Through a local git repository: task changes are found by a key in commit messages on the chosen branch, and in MR mode they come from the given Merge Request. No separate GitLab token is needed to fetch code.
Are Jira and MCP required?
No. Reviews run manually from an IDE without Jira, and MCP can be turned off.
Where does the source code go?
To the model you choose: a cloud provider or a local model connected in Codex, Cursor or OpenCode.
Which IDEs and agents are supported?
Codex, Cursor and OpenCode. The model is connected in the chosen environment, cloud or local.
What is in the final report?
A Markdown file `review-report-<task>.md` with findings, reasoning, fix suggestions and links to GitLab lines.
Discussion
No questions or reviews yet.
Related
All 1C:Enterprise toolsEarly access to an open platform for 1C development: export snapshots, a dependency graph, a local MCP, and edit drafts with versions and conflict checks
The 1c-dev Claude Code plugin: discussing a 1C task in dialogue, a technical spec, dependency-aware steps, and executing one verified step at a time
v8std standards harness
v8std-standards-harness
A skill and validator that stop an agent from silently skipping the v8std 1C development standards check on BSL code
A local read-only MCP server suite for 1C: a Rust code indexer, metadata skills, a prompt gallery and platform help search, with evidence-backed answers
claude_rules_1c: Claude Code rules and agents for 1C
claude_rules_1c
A set of 12 subagents, profile-based rules and MCP configuration for 1C development in Claude Code, with a three-layer rule loading system
A Claude Code and OpenCode plugin that will not let a session end until 1C edits pass standards, architecture and XML checks
README in short
The README tells about the two review stages, the auditor and critic roles, orchestration options and MCP usage. The quick start comes down to asking the agent to install RDV AIR, while manual installation and the Jira launch are moved to `docs/advanced-usage.md`.