Skip to main content
skill CX Codex

tendril-review — Claude Code skill: review every changed file

A skill that walks through all files touched by a big change, looking for dead code, duplication, leftover compatibility shims and test gaps.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add Ivy-Interactive/Ivy-Tendril --skill tendril-review -a claude-code
Open on GitHub

What tendril-review does

The skill starts from the git diff against your base branch. It gathers every changed file — staged, unstaged and committed — then sorts them into source, tests, config and prompt files. Anything that looks unrelated to the main change gets a question rather than a flag, because parallel manual edits happen.

Each source file is read in full and checked for dead code, unused imports, duplicated logic, methods that got too complex, unclear naming, missing error handling at system boundaries and abstractions that earn nothing. Backwards-compatibility code that may have outlived its purpose is listed separately. The skill never deletes it on its own; it asks first. Same for tests that no longer match the code.

The test suite runs through a command such as dotnet test — swap in your own. The report covers failures with a root-cause note, coverage gaps and tests that exercise behaviour you already removed. You get a structured summary at the end: files reviewed by category, issues with severity and location, cleanup candidates, and a prioritised action list. On a non-.NET project the default test command will not run, so replace it before you rely on that part.

What it helps with

  • review every file touched by a large change
  • spot dead code and unused imports
  • remove compatibility shims that are no longer needed
  • close test coverage gaps and drop obsolete tests
  • get a prioritised fix list before merging

Example prompts for the agent

  • Review the changes on this branch
  • Go through every modified file and tell me what can be cleaned up
  • Run the tests and flag the ones that are out of date

How to install tendril-review

1Skills CLI (Vercel)

npx skills add Ivy-Interactive/Ivy-Tendril --skill tendril-review -a claude-code

2Manually: put the files in place

~/.claude/skills/tendril-review/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/tendril-review/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about tendril-review

How do I install tendril-review in Claude Code?

Run npx skills add Ivy-Interactive/Ivy-Tendril --skill tendril-review -a claude-code. The skill lands in ~/.claude/skills/tendril-review/SKILL.md.

What does tendril-review do?

It reviews every file changed on the current branch for code quality problems, unnecessary backwards compatibility, missing tests and obsolete tests, then produces a report with recommendations.

Does the skill delete code by itself?

No. Backwards-compatibility code and obsolete tests are flagged, and the skill asks before removing anything.

Source text

SKILL.md · Apache-2.0
---
name: tendril-review
description: Comprehensive post-change review of all modified files. Checks for code smells, cleanup opportunities, unnecessary legacy support, missing tests, broken tests, and obsolete tests. Use after a big change to leave the codebase in better health.
---

# tendril-review

Perform a thorough review of all changes in the current working tree (or a specified branch/PR). The goal is to leave every touched file in better code health than before.

## Invocation

```
/tendril-review
```

No arguments required - operates on the current git diff against the base branch.

## What This Skill Does

1. Identifies ALL changed files (staged, unstaged, and committed on branch)
2. Reviews each file for code smells, cleanup opportunities, and quality issues
3. Flags unnecessary legacy/backwards-compatibility code (asks before removing)
4. Checks test coverage - missing tests, broken tests, obsolete tests
5. Runs the test suite and reports failures
6. Produces actionable recommendations

## Execution Steps

### Phase 1 - Scope the Changes

1. Run `git diff --name-only` against the base branch to get all changed files
2. Run `git status` to capture any uncommitted changes
3. Categorize files: source code, tests, config, promptware, other
4. Flag any files that seem unrelated to the main change - **ask the user** if there's confusion about why they changed

### Phase 2 - Code Quality Review

For each changed source file:

1. Read the full file content
2. Check for:
   - Dead code or unused imports
   - Code duplication
   - Overly complex methods (consider cyclomatic complexity)
   - Poor naming or unclear intent
   - Missing error handling at system boundaries
   - Inconsistent patterns vs. the rest of the codebase
   - Unnecessary abstractions or over-engineering
   - Legacy/backwards-compatibility code that may no longer be needed

**Important:** For any legacy support or backwards-compatibility code identified for removal - **ASK the user before removing**. Do not auto-delete.

### Phase 3 - Test Review

1. Identify tests related to changed code
2. Run the full test suite: `dotnet test` (or appropriate command)
3. Report:
   - **Failing tests** - investigate root cause
   - **Missing tests** - suggest what should be covered
   - **Obsolete tests** - tests that exercise removed/changed behavior and are no longer relevant
4. For obsolete tests - **ASK the user** before suggesting removal

### Phase 4 - Summary Report

Produce a structured report:

```
## Review Summary

### Changes Reviewed
- List of all files reviewed, grouped by category

### Issues Found
- [ ] Issue 1 - severity, file, description
- [ ] Issue 2 - ...

### Cleanup Opportunities
- [ ] Opportunity 1 - what and why

### Legacy Code Questions
- [ ] "This code appears to exist for backwards compat with X - still needed?"

### Test Status
- Passing: N
- Failing: N (with details)
- Missing coverage: list
- Potentially obsolete: list

### Recommendations
Prioritized list of actions to take
```

## Principles

- **Leave it better than you found it** - every touched file should improve in code health
- **Ask, don't assume** - when in doubt about whether something is still needed, ask
- **Unrelated changes are normal** - the user may have made parallel manual edits; clarify rather than flag as errors
- **Be thorough** - this is a big change review, not a quick scan

© Ivy-Interactive · Apache-2.0 · Ivy-Interactive/Ivy-Tendril

Similar skills

All skills for Code review and refactoring →

From the same set2

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor