1C ERP Diagnostics: the Gate 0-10 method
1C ERP Diagnostics
A ChatGPT and Codex plugin with an evidence-first 1C:ERP diagnostic methodology: eleven checkpoint gates from evidence intake to root cause
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
- Can propose production, accounting and access-rights changes classified as R3
- Works with executable 1C artifacts (CF, CFE, EPF) via isolated extraction and optional static analysis
Install
Manual install
Add the akim-kaneyev/1c-erp-diagnostics repository as a marketplace source in ChatGPT or Codex with git ref main, install the 1C ERP Diagnostics plugin as the primary entrypoint, and invoke it with @one-c-erp-diagnostics and your case description, or for Codex copy .agents/skills/one-c-erp-diagnostics/SKILL.md into the project and invoke $one-c-erp-diagnostics.
This is third-party code. Review the repository files before installing.
What it does
The plugin forces the agent to prove a causal chain: document, movement, register record, consuming mechanism, accounting result, symptom, rather than a plausible eyeball explanation. Eleven Gate 0-10 checkpoints sequentially discover available capabilities, inventory evidence with its provenance, build a specialist dependency graph, run independent read-only specialists in parallel where safe, adversarially challenge the preliminary conclusion at Gate 7, and validate the same analytics before and after a change. Every proposed action is classified by R0-R3 risk, and a final УСТАНОВЛЕНО (established) cause is only possible with a closed evidence chain and a passed adversarial review. It can optionally be installed as a marketplace of four independently maintained plugins: the diagnostic orchestrator itself plus Unica, 1C Skills in PowerShell, and 1C Skills in Python.
Who it is for. For people untangling complex 1C:ERP incidents (cost calculation, settlements, VAT, access rights) who want a verifiable cause instead of a plausible guess.
Good fit when
- The 1C:ERP incident is complex and a one-line explanation from the last visible symptom isn't enough
- You need a reproducible evidence chain from a document to the accounting result
- You need to compare two platform releases, or two similar documents, one correct and one problematic
Not a fit when
- You want a quick answer without formally going through the gates: the methodology is deliberately strict and multi-step
- You need to work with production .dt files, backups or broad confidential exports: the README explicitly forbids that
- The version as of this README, v0.3.9, is marked BLOCKED for runtime acceptance until all 26 synthetic cases pass in a fresh clean session
Example request
@one-c-erp-diagnostics figure out why the item's cost after month close differs from the calculated one, and give a proven causeLimitations
The README explicitly states that runtime acceptance of v0.3.9 is blocked until a full hashed run of all 26 synthetic cases passes in a fresh clean session, and the passing v0.3.8 run does not carry over to a changed package. This is an independent community project, not affiliated with 1C Company, OpenAI, Ingvar Consulting or the 1C Skills maintainers. Three companion marketplace plugins (Unica, 1C Skills PowerShell, 1C Skills Python) are pulled in as separate verified external sources and keep their own licenses; this repository does not copy them. There is no universal MXL parser; use an export to XLSX, XML, HTML or TXT instead.
How to disable. Remove the 1C ERP Diagnostics plugin from the ChatGPT/Codex marketplace, or delete the .agents/skills/one-c-erp-diagnostics folder for the repository-local skill.
Security check
- Can propose production, accounting and access-rights changes classified as R3
- Works with executable 1C artifacts (CF, CFE, EPF) via isolated extraction and optional static analysis
README in short
The English README describes why the project exists, an optional four-plugin marketplace ecosystem, eleven Gate 0-10 checkpoints with a purpose table for each, a list of 1C:ERP diagnostic domains, an evidence standard with source priority, a 0.3.3-0.3.9 version history with specific fixes to a strict machine-readable evaluation mode, extra capabilities like OpenSandbox and a pinned v8unpack, a quick start for the ChatGPT/Codex marketplace and the Codex skill, supported artifacts, and a safety and publication integrity section that explicitly forbids uploading production .dt files and credentials.
FAQ
Does the plugin change data in the 1C database itself?
Only under control: every action is classified by R0-R3 risk, and production, accounting or access changes fall under R3 and require explicit approval, rollback and Gate 9 validation.
What does УСТАНОВЛЕНО (established) mean?
The final cause status, allowed only with a closed source-evidence-premise-link-conclusion chain and a passed Gate 7 adversarial review, not a merely plausible explanation.
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