mj-agent-runtime-biz-catalog-sync

Detects drift between the qcm_catalog.yaml mirror and upstream naming standards, then proposes read-only diffs.

1|Updated Mar 4, 2026
One-click install
npx skills add https://github.com/MJ-AgentLab/mj-agent --skill mj-agent-runtime-biz-catalog-sync-mj-agentlab
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: mj-agent-runtime-biz-catalog-sync
Source: https://github.com/MJ-AgentLab/mj-agent/tree/main/.claude/skills/mj-agent-runtime-biz-catalog-sync
Command: npx skills add https://github.com/MJ-AgentLab/mj-agent --skill mj-agent-runtime-biz-catalog-sync-mj-agentlab

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When the upstream mj-system naming standard (Biz_DWS_Naming_Stability §2-§4) evolves—new metrics, renamed columns, deprecated entries—the mj-agent qcm_catalog.yaml mirror can silently drift, causing the find_biz_context tool to return wrong business semantics and the LLM to make incorrect decisions. This Skill detects that drift and proposes fixes without ever writing to the catalog directly. ## Core Features & Use Cases - Drift Detection: Wraps scripts/diff_biz_schema.py and scripts/fetch_biz_schema.py to compare qcm_catalog.yaml against the upstream standard and classify changes (new metric/period/dimension, renames, deprecations, semantic shifts). - Reverse Impact Scan: Greps dependent SKILL.md curated examples, golden_seed.jsonl reference SQL, system.md, and the SQL guardrail for references to renamed or removed catalog entries. - Read-only Proposals with HITL Gate: Outputs a draft diff plus impact analysis (Studio probe H1/H2/H3, smoke test effects) and stops for human approval—never edits qcm_catalog.yaml itself. - Use Case: After mj-system adds a new daily metric, run this Skill to get a proposed catalog diff, a list of affected curated NL-to-SQL examples, and review questions for the domain expert before any file is written. ## Quick Start Ask the agent to check biz catalog drift and sync qcm_catalog.yaml against the upstream mj-system naming standard.

Frequently Asked Questions about mj-agent-runtime-biz-catalog-sync

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

FAQPage Schema
How do I detect drift between a YAML catalog and an upstream naming standard?▼

Run the diff_biz_schema.py script via uv to compare qcm_catalog.yaml against the upstream standard, optionally fetching the latest upstream file first with fetch_biz_schema.py. The script reports new metrics, renamed columns, deprecated entries, and semantic shifts.

How do I find downstream files affected by a renamed metric or column?▼

Use grep-based reverse scanning across dependent SKILL.md curated examples, golden_seed.jsonl reference SQL, system.md, and the SQL guardrail module. Each hit is reported with file and line number, then classified as living (update to new name) or frozen (keep old name with a comment).

Can this Skill edit qcm_catalog.yaml directly?▼

No. It is read-only by design and never calls Edit or Write on the catalog file. It only proposes a draft diff and impact analysis; after human approval, a separate doc-authoring workflow performs the actual write.

When should I skip running a catalog drift check?▼

Skip it when changes are limited to YAML comments or file reformatting that do not alter metric, period, or dimension semantics. It is also not the right tool for editing SKILL.md, system.md, or SQL guardrail code, which have their own dedicated workflows.

Why does catalog drift cause wrong LLM answers in a data agent?▼

The find_biz_context tool retrieves business semantics from the catalog mirror at runtime. If the mirror lags the upstream standard, the LLM receives stale metric names or join keys and generates incorrect SQL or misleading business interpretations.