Claude
Skills
Sign in
Back

kiro-research

Included with Lifetime
$97 forever

Research AWS topics using Kiro CLI as an AWS expert (readonly mode). Use for: AWS documentation, error investigation, CloudFormation/CDK questions, AWS best practices. Note: Uses Kiro's training knowledge only (no web search).

Cloud & DevOps

What this skill does


# Kiro Research Skill

This skill provides integration with Kiro CLI for AWS research and troubleshooting tasks.
Kiro runs in **readonly mode** - answering from its trained AWS expertise without web search.

## Prerequisites

- Kiro CLI installed (visit https://kiro.dev or download from AWS)
- AWS credentials configured (for AWS-specific operations)

## Available Scripts

All scripts are located in `${CLAUDE_PLUGIN_ROOT}/scripts/`.

### 1. Check Installation

```bash
${CLAUDE_PLUGIN_ROOT}/scripts/check-kiro.sh
```

Verifies Kiro CLI is installed and accessible.

### 2. Research Mode (kiro-cli chat)

```bash
${CLAUDE_PLUGIN_ROOT}/scripts/kiro-ask.sh "your research question"
```

Use for:
- AWS service documentation lookup
- CloudFormation/CDK troubleshooting
- AWS error investigation
- Best practices inquiry
- Architecture recommendations

Example:
```bash
${CLAUDE_PLUGIN_ROOT}/scripts/kiro-ask.sh "How do I configure VPC endpoints for S3?"
```

## Workflow

1. First, verify Kiro CLI is available using `check-kiro.sh`
2. For research tasks, use `kiro-ask.sh` with the research prompt
3. Summarize findings and provide actionable recommendations

## Kiro CLI Options Reference

| Option | Description |
|--------|-------------|
| `--no-interactive` | Run without user input (used by scripts) |
| `--agent <AGENT>` | Specify agent to use |
| `--model <MODEL>` | Specify model to use |
| `-a, --trust-all-tools` | Allows model to use any tool without confirmation |
| `--trust-tools=<TOOL_NAMES>` | Trust only specific tools (e.g., `--trust-tools=fs_read,fs_write`) |

## Error Handling

- **Timeout**: Commands timeout after 120 seconds
- **Not installed**: Provides installation instructions
- **Credentials**: Prompts to configure AWS credentials if needed

## Input Sanitization

For security, prompts are sanitized before execution:
- Newlines and carriage returns are removed (multi-line prompts become single-line)
- Backticks (`) and dollar signs ($) are stripped to prevent prompt injection

## Notes

- This skill uses `context: fork` to isolate large outputs from the main conversation context.
- **Readonly mode**: Kiro answers from its training knowledge only (no web search or external tools).
- For web search needs, use Claude's built-in WebSearch or Gemini instead.

Related in Cloud & DevOps