Pi 1C Agent: a 1C profile for Pi Coding Agent
Pi 1C Agent
A Pi Coding Agent profile for 1C development: ASK, PLAN and BUILD modes, specialized subagents, per-configuration knowledge and optional MCP tools
High risk
We rate an entry high when the tool writes to external systems, handles money, production databases or secrets, or runs arbitrary commands. The CLI installs it only with your consent.
Why this level
- In BUILD mode the agent changes source files, builds artifacts and can load the configuration into a test infobase via /update1cbase
- The optional memory and browser MCPs send project data to external or local services once separately enabled
Install
Manual install
git clone https://github.com/z03ps00/pi-1c-agent.git ~/pi-1c-agent
export PI_CODING_AGENT_DIR=~/pi-1c-agentCloning the profile on Linux and macOS.
This is third-party code. Review the repository files before installing.
What it does
The profile adds a layer for real 1C development on top of Pi Coding Agent: ASK mode only reads and answers, PLAN researches a task and drafts a plan without writing code, and BUILD changes files and calls tools. Before any change, the agent decides whether to write into the main configuration or into an existing extension. Thirteen specialized subagents (analyst, explorer, planner, developer, tester, code and architecture reviewers, a defect fixer and others) can run in parallel for read-only roles and sequentially for ones writing to the same tree. Knowledge about a specific configuration lives in .pi/1c with evidence and requires confirmation before activation, and MCP tools (a 1C Docker bundle, EDT MCP, Vanessa, memory, browser) are connected separately through the /installtools wizard. Commands exist for exporting from an infobase, building .cf/.cfe artifacts and a test-deploy cycle.
Who it is for. For 1C developers who already use Pi Coding Agent and want a specialized profile with modes, subagents and per-configuration knowledge.
Good fit when
- You need to explore an unfamiliar or complex 1C configuration in read-only mode before changing it
- You need a change plan with risks and a verification approach before the agent starts writing code
- You need separate subagents for analysis, development, testing and review, running in parallel where it's safe
Not a fit when
- You use an agent other than Pi Coding Agent: the ASK/PLAN/APPROVE runtime protections are Pi-specific and are not reproduced one to one elsewhere
- You want a simple one-off script without the ASK/PLAN/BUILD model and subagents
- The 1C platform is not installed and there's no access to the database to work with an infobase
Example request
/mode plan I need to add a stock check before posting a shipment document, study the current implementation and propose a planLimitations
The profile is built for Pi Coding Agent, a third-party project not on SkillFoxx's agent list: the README states Cursor reads AGENTS.md as plain instructions, but the ASK/PLAN/APPROVE mode protections are a Pi-only mechanism and don't work the same way in Cursor. MCP services the profile knows about are not considered installed automatically; they're connected separately through /installtools. Some material is adapted from third-party repositories (comol/ai_rules_1c, Desko77/cursor-1c-skills, comol/Humanizer_RU), with exact license boundaries listed in NOTICE. Requires Node.js 22.19 or newer and Pi Coding Agent 0.85.x installed.
How to disable. Remove the PI_CODING_AGENT_DIR variable or the cloned profile folder so Pi stops using it.
Security check
- In BUILD mode the agent changes source files, builds artifacts and can load the configuration into a test infobase via /update1cbase
- The optional memory and browser MCPs send project data to external or local services once separately enabled
README in short
The README describes in detail a Pi Coding Agent profile for 1C: a feature table, ASK/PLAN/BUILD modes with a mermaid task-flow diagram, a list of thirteen specialized subagents, commands for working with a test database and release artifacts, a project-knowledge mechanism with context priority, session rotation on context fill, opt-in MCP connection through separate installers, a skills list including Humanizer RU and КД 2/3, approval modes, the repository layout, an important-limitations section, and license credits to three third-party projects that parts of the profile build on.
FAQ
Does the profile work the same way in Cursor as in Pi?
No, the README explicitly warns that Cursor reads AGENTS.md as instructions, but the mode and /approve runtime protections are specifically a Pi-runtime mechanism.
What happens if the session context fills up during a long task?
With /session-rotate enabled, the agent writes a handoff with the task state, verifies it, and continues in a new Pi session linked to the previous one.
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