What problem does it solve? Before writing a Plan, SPEC, or code for the mj-agent data agent, teams need to confirm that the Issue and Plan still match the actual repository state. This Skill performs a read-only fact-check across code, configuration, business catalog, tests, and documentation so downstream stages start from verified evidence rather than stale assumptions. ## Core Features & Use Cases - 8-dimension repository scan: Covers the seven mj-agent source modules, LangGraph Studio and Chainlit surfaces, real biz data flow, databases, configuration and secrets, tests, and documentation governance. - Reverse stale-doc scan: Greps docs, SKILL.md files, and prompts for renamed functions, moved files, and renamed SQL objects, plus biz_catalog drift detection via scripts/diff_biz_schema.py. - Documentation Decision matrix: Produces a 10-row Create/Update/None decision table across Plan, SPEC, ADR, RUNBOOK, GUIDE, STANDARD, ISSUE, ASSESSMENT, CHANGELOG, and INDEX document types. - Use Case: A developer says "按 plan 实施" on an existing branch; the Skill scans the worktree, confirms the Plan still holds, escalates risk if runtime SKILL.md or prompt versions changed, and outputs a Repo Scan Result with HITL questions before any code is written. ## Quick Start Ask the assistant to run a repo scan against the current branch and Issue before starting implementation, for example by saying "开始执行，先做一次 repo scan".
