Files
agent-skills/skills/code-review-agent/SKILL.md
2026-08-05 15:46:16 +00:00

77 lines
2.2 KiB
Markdown

---
name: code-review-agent
version: 1.0.0
description: Automate code review process using a multi-step workflow with human-in-the-loop
approval.
inputs:
- Repository diff or code changeset (string)
- User ID (string)
steps:
- 'Step 1: Build the code review graph using `build_graph()` from `agentkit.workflow.code_review.graph`'
- 'Step 2: Invoke the graph with initial parameters including the repository diff
and user ID, and set thread_id as a configurable parameter'
- 'Step 3: The graph processes the input through various steps until completion or
human approval is needed'
outputs:
- Review result (dictionary containing messages, issues, etc.)
tags: []
metadata:
source_repo: https://github.com/itszhaoziyan-n/AgentKit.git
extracted_at: ''
confidence: 0.95
---
# code-review-agent
Automate code review process using a multi-step workflow with human-in-the-loop approval.
## Setup
**Dependencies:**
```text
pip install langgraph>=0.3 langchain-core>=0.3 langchain-anthropic>=0.3 langfuse>=2.0 mcp[server]>=1.24,<2.0 langchain-mcp-adapters>=0.1 tenacity>=9.0
```
**Setup steps:**
1. cp .env.example .env
1. docker compose up -d
1. pip install -e '.[dev]'
## Key Files
- `agentkit/workflow/code_review/graph.py - Defines the code review graph and its invocation method.`
## Steps
1. Step 1: Build the code review graph using `build_graph()` from `agentkit.workflow.code_review.graph`
2. Step 2: Invoke the graph with initial parameters including the repository diff and user ID, and set thread_id as a configurable parameter
3. Step 3: The graph processes the input through various steps until completion or human approval is needed
## Implementation Details
```python
graph = build_graph()
thread_id = str(uuid.uuid4())
result = graph.invoke(...)
```
## Inputs
- Repository diff or code changeset (string)
- User ID (string)
## Outputs
- Review result (dictionary containing messages, issues, etc.)
## Failure Modes
- Specific failure scenario with mitigation: If the graph invocation fails due to an unexpected state, it will halt and require manual intervention.
## Source
Extracted from: [https://github.com/itszhaoziyan-n/AgentKit.git](https://github.com/itszhaoziyan-n/AgentKit.git)
Confidence: 0.95