---
name: ry-repair
description: "Нормализация репозитория: source-of-truth scan, semantic entropy audit, repair plan, technical-only fixes, validators, docs/memory sync. Используй для: /rldyour-flow:ry-repair, почини систему, нормализуй репозиторий, убери противоречия, repair repo. EN triggers: repository repair, semantic entropy cleanup, contract normalization, stale docs repair, AI-tool context repair."
argument-hint: "<scope or problem>"
---

# ry-repair

## Purpose

Normalize a repository so Claude Code, Codex, OpenCode, Gemini CLI, and MiMoCode can operate from the same verified facts with minimal semantic entropy. This is a technical repair flow, not a permission to change business logic silently.

For rldyour AI CLI configuration repositories, `/ry-repair` also verifies deterministic install/update/sync convergence through the root `config/ry-repair-sync-contract.json` and `scripts/ry_repair_sync.py` contract when that control plane is present.

## Mode Selection

- **Work project**: run the full authoring repair workflow below.
- **rldyour control plane or adapter config repo, goal = converge the installed
  system**: run consumer mode instead - update the checkout FROM GitHub
  (`git pull --ff-only`, `git submodule update --init --recursive`), then
  `python3 scripts/ry_repair_sync.py --check` and `--apply-system`.
  In consumer mode do NOT author changes into the repository: no commits, no
  doc/contract/memory edits, no fullrepo publication. If fast-forward fails or
  validators reveal repository drift, report it and switch to authoring repair
  only on explicit owner instruction.

## Workflow

1. Detect repository type, active branch/worktree, submodules, CI surface, and effective project policy from `RLDYOUR_PROJECT_POLICY`, `.rldyour/project-policy.local.json`, `.rldyour/project-policy.json`, or built-in defaults.
2. Read project instructions (`AGENTS.md`, `.claude/CLAUDE.md`, OpenCode/Codex/Claude config when present) and identify the declared sources of truth.
3. Inspect Serena memories, plans, and research archives for stale facts, unsupported claims, missing taxonomy, or contradictions with current code.
4. Inspect GitHub issues/PRs/history when available through MCP/app/CLI, then verify every issue against current code before treating it as a fact.
5. Inspect MCP/LSP/tooling config, hook lifecycles, commands/skills/agents, CI gates, release manifests, and dependency baselines.
6. When the root control plane is present, run `python3 scripts/ry_repair_sync.py --plan --target "$PWD"` and use `--check` before claiming local repo, system AI CLI config, Serena memory, or GitHub/fullrepo sync. For OS/mode work, pass explicit flags: `--os macos|linux|wsl|windows`, `--mode standard|orchestrator`, and `--cmux` only for macOS orchestrator mode.
7. Treat approved active inventories as system-wide: installed Claude/Codex/OpenCode configs, active agent/tool surfaces, CI workflows, runtime pins, docs, and release gates must match current MCP/provider policies. Do not keep permanent tool-specific absence gates for removed components.
8. Detect semantic entropy: duplicated docs, stale pins, conflicting instructions, dead config, unclear source-of-truth, missing ADR/CONTEXT/FUTURE facts, broken validator contracts, and natural-language policy overrides that are not materialized in project policy JSON.
9. Produce a repair plan that separates:
   - technical repairs the agent may apply;
   - business, functional, security-posture, deployment-target, data-model, or ADR decisions that require the owner.
10. Ask the owner in Russian before changing any decision-class item. For Claude Code, use AskUserQuestion when available; otherwise present concise options with a recommendation and impact.
11. Apply technical-only repairs using existing project patterns and native adapter surfaces. If the user explicitly states repository policy and no policy file exists, write `.rldyour/project-policy.json` or `.rldyour/project-policy.local.json` only from that explicit instruction; do not infer ownership/destructive permissions from chat tone.
12. Run matching validators, tests, schema checks, hook smoke, release/archive checks, installed-config checks, and instruction/memory freshness checks.
13. Synchronize durable docs and Serena memories from verified code/config state, then finish through `flow-post-task-sync` when the repair changed durable artifacts.

## Non-Negotiables

- Code, current config, current runtime checks, and verified GitHub state are the source of truth. Memories and docs are derived facts.
- Do not edit ADR meaning, business logic, functional behavior, pricing, deployment targets, security posture, or data contracts without owner approval.
- Do not make hooks perform long-running repair work. Hooks may mark state; `ry-repair` performs the repair.
- Do not hide unresolved drift behind "green" summaries. Every blocked check must name the blocker and next proof command.
- Do not report local repo, system config, Serena memory, or GitHub parity without evidence from `ry_repair_sync.py`, installed-config validators, current git state, and fullrepo/GitHub checks where applicable.

## Output

Report in Russian:

- Scope and source-of-truth map.
- Confirmed drift, grouped by severity.
- Technical repairs applied.
- Decision-class items left for owner approval.
- Exact validation commands and results.
- Docs/memory/fullrepo/git synchronization status.