skills.rest
pavelvdo
pavelvdoCOMMUNITY
Explore all skills by this author →
GitHub

1c-agent-patterns

Guides delegation of 1C development tasks to specialized subagents with role-based prompt templates.

Updated Jun 26, 2026
Search Tags:#code review#agent delegation#subagents#openspec#prompt templates#1c#bsl
One-click install
npx skills add https://github.com/pavelvdo/universal-xml-exchange2 --skill 1c-agent-patterns-pavelvdo
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: 1c-agent-patterns
Source: https://github.com/pavelvdo/universal-xml-exchange2/tree/main/.cursor/skills/1c-agent-patterns
Command: npx skills add https://github.com/pavelvdo/universal-xml-exchange2 --skill 1c-agent-patterns-pavelvdo

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Coordinating multiple specialized AI agents for 1C:Enterprise development is error-prone: choosing the wrong agent, writing inconsistent prompts, or skipping quality gates leads to poor code and wasted iterations. This Skill provides a navigator for delegating 1C tasks to role-based subagents (architect, writer, reviewer, explorer, simplifier, quality-controller) with ready-made prompt templates and shared instruction blocks. ## Core Features & Use Cases - Complexity assessment and agent routing: Classify tasks (simple/medium/complex/critical) and select the right combination of explorer, architect, writer, and reviewer agents. - Role-based prompt templates: Per-role files (architect.md, writer.md, reviewer.md, explorer.md, simplifier.md, quality-controller.md) with copy-paste Task() call templates for design, implementation, code review, bug fixes, slice decomposition, and verification layers. - Shared instruction blocks (gates): Reusable HALT-gates such as DATA_CONTRACT_GATE, EXTENSION_GUARD, and EXISTING_MECHANISM_GATE that enforce 1C coding discipline across all agent prompts. - Use Case: When implementing a new feature in a 1C configuration, use this Skill to assess complexity, dispatch an explorer to map existing modules, have an architect produce design.md, delegate BSL coding to the writer with contract gates, and run a mandatory reviewer pass with auto-fix iterations. ## Quick Start Ask the assistant to plan agent delegation for your 1C development task using the 1c-agent-patterns reference, starting with a complexity assessment.

Frequently Asked Questions about 1c-agent-patterns

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I delegate 1C coding tasks to AI subagents?▼

Assess task complexity first, then call the Task tool with the appropriate subagent_type: onec-code-explorer for research, onec-code-architect for design, onec-code-writer for BSL implementation, and onec-code-reviewer after every writer pass. Use the per-role prompt templates in architect.md, writer.md, reviewer.md, and explorer.md.

What are shared instruction blocks in 1C agent prompts?▼

Shared instruction blocks are named reusable gate texts (DATA_CONTRACT_GATE, INTEGRATION_CONTRACT_GATE, EXISTING_MECHANISM_GATE, EXTENSION_GUARD, CONTEXT_SAFETY, CONDITIONAL_ACTION_GATE, ORCHESTRATOR_IMPLEMENTATION_GATE) that the orchestrator copies into agent prompts. They enforce HALT-checks against defensive coding, extension corruption, and duplicated logic.

How is task complexity assessed for 1C agent delegation?▼

Complexity is rated by file count and architectural impact: simple (1-2 files, one explorer/writer/reviewer), medium (3-5 files, adds architect), complex (5+ files with deep analysis), and critical (architectural changes with multisampling across 2-4 explorers and 2 architects).

What should I do when a 1C subagent call fails with Invalid enum value?▼

First verify the tool name is Task and the subagent_type spelling is correct, then retry once. If the retry fails, stop and report the failing agent, exact error, and attempts to the user. Never fall back to a generalPurpose agent, since that masks the error and lowers result quality.

When are writer and reviewer agents not used in 1C workflows?▼

In explore mode (opsx:explore) only analytical agents run: trace-analyst, explorer, and architect. Implementation agents (writer, reviewer) are not launched during exploration; they activate only in the apply phase after a verified design exists.

Related Skills

Explore More Software Engineering Skills→
darkroomengineeringdarkroomengineering
41

fix

Investigate and resolve software bugs through a structured six-stage workflow.

Official
Advanced
ReadyStateChangeReadyStateChange

slicing-work

Break shaped plans into vertical, demoable slices with sequencing and dependencies.

Official
Intermediate
Desko77Desko77
40

1c-epf-full-cycle

Orchestrate the complete EPF creation workflow for 1C external processing.

Community
Advanced
LongTermSupportLongTermSupport

phpunit-runner

Execute PHPUnit test suites and delegate failure fixes to agents.

Official
Intermediate