# Update Setup

> >-

- **Type:** Skill
- **Install:** `agentstack add skill-alexandrbasis-claudops-update-setup`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [alexandrbasis](https://agentstack.voostack.com/s/alexandrbasis)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [alexandrbasis](https://github.com/alexandrbasis)
- **Source:** https://github.com/alexandrbasis/claudops/tree/main/.claude/skills/update-setup

## Install

```sh
agentstack add skill-alexandrbasis-claudops-update-setup
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Update Setup

> **Announcement**: "Checking for upstream workflow updates from **claudops**..."

Sync local `.claude/` workflow files with upstream claudops. The skill is
upstream-driven: local-only user files are ignored unless they are explicitly tracked in
the update manifest.

## Hard Rules

- Start with the deterministic script. Do not classify diffs manually.
- Ask for explicit approval before applying updates, refreshing disabled files, deleting
  removed upstream files, branch creation, commit, push, or PR creation.
- Never update `.claude/settings.json`, `.claude/settings.local.json`, `CLAUDE.md`,
  hook logs, `.gitkeep`, or local-only `*.local.*` files through this skill.
- Use LLM judgment only to explain conflicts and help the user choose a strategy after
  the script has produced exact statuses and diffs.

## Script

Use:

```bash
python3 .claude/skills/update-setup/scripts/update_setup.py --help
```

The script writes adoption state to:

```text
.claude/skills/update-setup/claudops-upstream.lock.json
```

Tracked scope:

- `.claude/**`

Skipped scope:

- `.claude/settings.json`
- `.claude/settings.local.json`
- `.claude/hooks/logs/**`
- `.gitkeep`
- `CLAUDE.md`
- local-only `*.local.*`

## Workflow

### 1. Scan

Create a deterministic report:

```bash
python3 .claude/skills/update-setup/scripts/update_setup.py scan \
  --output .claudops-update-report.json
```

For local testing against an existing upstream clone:

```bash
python3 .claude/skills/update-setup/scripts/update_setup.py scan \
  --upstream-root /tmp/claudops-upstream-sync \
  --commit "$(git -C /tmp/claudops-upstream-sync rev-parse HEAD)" \
  --output .claudops-update-report.json
```

### 2. Present Report

Render the report:

```bash
python3 .claude/skills/update-setup/scripts/update_setup.py report \
  --report .claudops-update-report.json
```

Explain statuses:

- `new`: upstream file is not present locally.
- `modified`: upstream differs from local and local has no tracked conflict.
- `conflicting`: local changed since the last tracked adoption.
- `disabled`: upstream file exists, but local has `.disabled`.
- `removed`: file was tracked in the manifest but is gone upstream.
- `placeholder_only`: differences are filled `{{PLACEHOLDER}}` values.
- `unchanged`: local matches upstream or the tracked adoption hash.

If only `unchanged` and `placeholder_only` entries exist, report that the setup is up to
date and stop.

### 3. Ask for Approval

Ask which exact paths to update, refresh, or delete. Paths are relative to `.claude/`.
Convert approval into a selection JSON file:

```json
{
  "update": [
    "skills/example/SKILL.md"
  ],
  "refresh_disabled": [
    "skills/disabled/SKILL.md"
  ],
  "delete": [
    "skills/removed/SKILL.md"
  ]
}
```

Rules:

- `update` may include `new` and `modified` paths.
- `refresh_disabled` may include `disabled` paths and writes `.disabled`.
- `delete` may include only `removed` paths.
- Do not apply `conflicting` paths until the user chooses a conflict strategy.

### 4. Apply

After approval:

```bash
python3 .claude/skills/update-setup/scripts/update_setup.py apply \
  --report .claudops-update-report.json \
  --selection /path/to/selection.json
```

The apply step updates selected files and refreshes the lock manifest.

### 5. Verify

Run:

```bash
python3 .claude/skills/update-setup/scripts/update_setup.py verify
```

When changing this skill itself, also run:

```bash
python3 .claude/skills/update-setup/tests/test_update_setup.py
```

## Conflict Handling

For `conflicting` files:

1. Show the file path and exact diff from `.claudops-update-report.json`.
2. Explain what upstream changed and what local changed.
3. Ask the user to choose one strategy:
   - replace local with upstream
   - keep local and mark skipped
   - manually merge, then update the manifest after verification

Do not silently merge conflicts.

## Post-Update Checks

After applying selected updates:

- Run `verify`.
- If new hook files were added, inspect `.claude/settings.json` and report unwired hooks
  as manual follow-up. Do not auto-edit settings.
- If updated files contain `{{PLACEHOLDER}}`, tell the user to run `/setup`.
- Summarize counts: updated, disabled refreshed, deleted, skipped, placeholders, and
  manual follow-ups.

## Edge Cases

- Clone failure: report the git error and stop.
- Missing `.claude/`: tell the user to clone the workflow repo first or run `/setup`.
- More than five `conflicting` files: warn this is a major update and suggest reviewing
  upstream commits before applying.
- Binary files: the script may copy them, but conflict explanation should avoid trying to
  summarize binary content.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [alexandrbasis](https://github.com/alexandrbasis)
- **Source:** [alexandrbasis/claudops](https://github.com/alexandrbasis/claudops)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-alexandrbasis-claudops-update-setup
- Seller: https://agentstack.voostack.com/s/alexandrbasis
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.