comol/ai_rules_1c/mcp-1c-tools
Claude Code and agent skill by comol/ai_rules_1c
- community
Checking that the SKILL.md file is still in the repository.
Router for non-memory 1C MCP operations: select the server, operation skill with exact arguments, and fallback. Load before choosing a 1C tool; memory-only work enters through project-memory.md.
Install
npx vskill@latest install comol/ai_rules_1c/mcp-1c-toolsSKILL.md
MCP tools for 1C — router
Apply content/rules/mcp-policy.md → Tool availability before routing: .dev.env TOOL_* policy plus callable tools and verified scope. Config entries alone prove nothing. The same policy owns obligations, budgets and typed-error recovery. Search: content/rules/mcp-first-search.md; task sequences: content/rules/tooling-playbooks.md.
Project scope is part of the call. For extensions or a multi-project graph, first match the current source roots to a returned base project_id through list_graph_projects (content/rules/multi-contour-search.md). Supply it explicitly on every graph project-data call that exposes it, including paging and evidence. Resolve extension-layer coverage separately. Use each other server's actual selector (configurationId or another field only if exposed); its identifiers are not interchangeable with graph IDs. If a tool has no explicit selector, require a verified fixed/session/entity scope or use a suitable scoped tool/fallback. Never invent unsupported parameters or accept a default/foreign project as the current one. Reuse a valid mapping until the workspace/server context changes.
Need → operation skill
Load the skill for the operation, not this whole catalogue. Each skill lists the exact argument names of its tools and example calls; no schema fetch is needed for a tool it names.
| Need | Skill | Servers | |
|---|---|---|---|
| Locate or read BSL code, module layout, members of a context | content/skills/1c-code-search/SKILL.md | graph, code | |
| Facts about a metadata object: passport, attributes, tabular-part columns, objects by category or description | content/skills/1c-meta-info/SKILL.md | graph, code | |
| Usages, call chains, downstream impact, register writers, extension layers | content/skills/1c-impact/SKILL.md | graph, code | |
| Read forms, form artifacts, XSD / format specs before a form change | content/skills/1c-form-inspect/SKILL.md | code, graph, docs | |
| Validate changed BSL and metadata XML (Gates 1–3, 5) | content/skills/1c-validate/SKILL.md | syntax, checker, code | |
| Platform reference, capability check, БСП API, routed standards, ITS, configuration docs | content/skills/1c-platform-help/SKILL.md | docs, ssl, checker, code | |
| Templates as the base | content/skills/1c-templates-memory/SKILL.md | templates | |
| Memory recall / save (independent entry point) | content/rules/project-memory.md | cognee, openviking, templates | |
| Run a query or fragment in the live infobase, last event-log error | content/skills/1c-live-ib/SKILL.md | data | |
| Check behaviour in the 1C interface (thin client): forms, fields, tables, commands, messages | content/skills/1c-qa-testing/SKILL.md + content/rules/qa-testclient.md | qa | |
| Create / edit / remove metadata, forms, roles, DCS, MXL, infobases | content/skills/1c-metadata-manage/SKILL.md | scripts, not MCP | |
Live 1C:EDT workspace (USE_EDT=true only) | docs/edt-mcp.md, content/rules/edt-workflow.md | edt |
Server catalog
The operation table above is sufficient for ordinary routing. Read server-catalog.md only to identify an unfamiliar server/alias, find a server-specific reference, or select an optional pre-alpha integration. Do not load the whole catalog before a known operation.
Fallback chain
Project source (code, metadata, usages, forms, file locations): within verified contour coverage, graph → mapped code-metadata → scoped native Grep / Glob / Read after a bounded miss, with a one-line fallback note. Code chooses its file-scan fallback internally; current tools have no grep input. Skip uncovered lanes; no eligible exposed index means native search in that contour immediately. Owner: content/rules/mcp-first-search.md; multiple roots, catalog/scope selectors and acceptance: content/rules/multi-contour-search.md.
External knowledge has no interchangeable fallback chain. Select only the relevant eligible provider. Missing templates: disclose no template check; memory: project-memory.md; routed standards: help-corpus-retrieval.md; validators/live IB: verification-gates.md. An unavailable source never proves API absence or a passed check; block dependent design when authoritative facts are missing.
Open docs/<server>.md once per server per session and only for a mode or response shape the operation skill does not cover; the environment descriptor wins over any document here.
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingcontent/skills/mcp-1c-tools/SKILL.md
Popularity
- 7-day trend score
- 5
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/comol/ai_rules_1c/mcp-1c-tools)