1c-ai-skills
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
Low risk
We rate an entry low when it mostly gives the agent instructions and reference material.
Why this level
- The plugin explicitly does not load the configuration into the database or write data, it only prepares a spec and code for manual application
- The evals commands run paid model calls via claude plugin eval if a developer chooses to use them
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/1c-ai-skillsDetects the agents on your machine, checks the risk and pins the version.
Other ways to install
Run one by one in the Claude Code chat
/plugin marketplace add blessed2k/1c-ai-skills
/plugin install 1c-dev@1c-ai-skillsIn Claude Code run /plugin marketplace add blessed2k/1c-ai-skills and /plugin install 1c-dev@1c-ai-skills, restart the session, then run /1c-dev:setup in the project folder and drive the task through the grill, spec, tickets, step cycle.
Other ways from the author
/plugin marketplace add blessed2k/1c-ai-skills
/plugin install 1c-dev@1c-ai-skillsAfter installing, restart the session and run /1c-dev:setup in the project folder.
This is third-party code. Review the repository files before installing.
What it does
The 1c-dev plugin ships nine skills. Four drive the task cycle: grill interviews the user in rounds until a shared understanding of the solution emerges, gathering configuration facts itself; spec turns the discussion into a technical spec with metadata, register movements, the insertion point, permissions and verification; tickets splits the spec into dependency-aware steps with an assignee, after each of which the database stays working; step takes the first free step, does it, verifies it and stops. Five reference skills the agent invokes itself when it recognizes the situation: BSL standards from ITS, a module skeleton with the standard region structure, a YAxUnit test scaffold, a Vanessa Automation pitfalls reference, and project setup for the plugin. At no stage does the agent load the configuration into the database or write data itself; it only prepares changes and tells the person what to do.
Who it is for. For 1C teams on Claude Code who want the agent to first agree on a solution and write it down, rather than jumping straight into guessing code.
Good fit when
- You want a structured process from task discussion to step-by-step implementation with verification at each step
- The agent keeps making changes in the wrong place: duplicating code across modules instead of a shared procedure
- You need decisions remembered between sessions as a spec file you can approve and return to
Not a fit when
- You need an agent that edits XML exports and builds objects itself: 1c-dev decides what and where, it does not perform the edits, the authors recommend pairing it with cc-1c-skills for that
- You have no configuration XML export in the project: the skills still work, but the spec comes out without fact-checking against the base
Example request
I need to add a new discount type to the Order document, start by discussing the solutionLimitations
Guaranteed to work only in Claude Code: slash commands like /1c-dev:spec and the restriction on the agent invoking a skill on its own are tied to its behavior; the format can be copied into other SKILL.md-aware agents, but behavior is not guaranteed there. The full text of ITS standards requires an its.1c.ru subscription; the skills only give a paraphrase with the standard number. The verification evals run via claude plugin eval and are paid model runs billed to the user's account.
How to disable. Run /plugin uninstall 1c-dev@1c-ai-skills in Claude Code and, if needed, remove the marketplace with /plugin marketplace remove 1c-ai-skills.
Security check
- The plugin explicitly does not load the configuration into the database or write data, it only prepares a spec and code for manual application
- The evals commands run paid model calls via claude plugin eval if a developer chooses to use them
README in short
The README explains the motivation: an agent working on 1C more often errs in where to change code than in syntax, so the skills first lock in a decision and only then do the work in small verifiable steps. It gives install steps via /plugin marketplace add and install, tables of all nine skills with descriptions, a grill-spec-tickets-step cycle diagram, a settings section (edit mode, authorship marker, verification commands, MCP server) and requirements, including a recommendation to pair it with cc-1c-skills for XML edits. A development section lists Node.js repo-verification scripts and evals for behavioral testing. The cycle skills grew out of mattpocock/skills. MIT license.
FAQ
Does the agent load changes into the database itself?
No, at no stage of the cycle does the agent load the configuration, update database structure or write data, it only prepares changes and reports what a person should do.
How is 1c-dev different from cc-1c-skills?
1c-dev decides what and where to change through the discussion and spec cycle, while cc-1c-skills performs the actual XML export edits; the authors recommend using them together.
Related
A skills library that gives coding agents a development process: brainstorming, planning, TDD, subagents and code review
Skills for real engineers by Matt Pocock
Skills For Real Engineers
Small composable skills for engineering with agents: plan grilling, TDD, bug diagnosis, code review and architecture
GitHub toolkit for spec-driven development: the specify CLI adds agent commands and skills to a project, from principles to implementation
Reference MCP servers
Model Context Protocol servers
Official reference MCP servers: Filesystem, Fetch, Git, Memory, Sequential Thinking, Time and Everything