メニュー
Invoked when deploying OpenContext to AWS, setting up Lambda infrastructure, provisioning a new environment, or running opencontext deploy. Use when the user says "deploy", "set up AWS", "push to Lambda", "provision staging/prod", or "connect Claude to my deployment".
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
Invoked when the user wants to create a new plugin, add a new data source, extend OpenContext with a custom integration, or implement a new MCPPlugin or DataPlugin subclass in custom_plugins/. Use when the user says things like "add a plugin", "create a custom plugin", "integrate with [service]", or "implement a new data source".
Invoked when test coverage is below 80%, CI fails on the coverage gate, or the user asks to improve test coverage for a module. Use when the user says "coverage is failing", "CI is failing on tests", "increase coverage", or "add tests for X".
Invoked when diagnosing issues with a deployed Lambda function, investigating CloudWatch errors, or troubleshooting MCP tool failures in a staging or production environment. Use when the user says "Lambda is failing", "MCP tools aren't working", "production errors", or "debug my deployment".
SOC 職業分類に基づく
| name | deploy-aws |
| description | Invoked when deploying OpenContext to AWS, setting up Lambda infrastructure, provisioning a new environment, or running opencontext deploy. Use when the user says "deploy", "set up AWS", "push to Lambda", "provision staging/prod", or "connect Claude to my deployment". |
| command | /deploy-aws |
source .venv/bin/activate (local virtual environment is activated)config.yaml exists with exactly one plugin enabledaws configure or env vars)uv sync --all-extras already runuv run opencontext validate --env {env}
Checks config.yaml structure + Terraform syntax. Fix any errors before proceeding.
uv run opencontext security
Runs pip-audit internally before every deploy path; fix any HIGH/CRITICAL findings first.
uv run opencontext configure
Interactive: creates terraform/aws/{env}.tfvars, initializes Terraform workspace {city}-{env}.
Workspace naming: {city}-{env} — e.g., chicago-staging, boston-prod.
uv run opencontext deploy --env {env}
This packages the Lambda ZIP (validates <250 MB), applies Terraform, outputs the API Gateway URL.
uv run opencontext status --env {env}
uv run opencontext test --url {api_gateway_url}/mcp
Copy the API Gateway URL → Claude Desktop → Settings → Connectors → Add MCP Server.
| Symptom | Cause | Fix |
|---|---|---|
ConfigurationError: 0 or 2+ plugins enabled | config.yaml misconfigured | Check one-plugin rule |
| Lambda timeout on cold start | lambda_timeout too low | Increase in config.yaml (max 900s) |
| Package size exceeded | Dependencies too large | Run opencontext validate to check size |
| Terraform state conflict | Concurrent apply or stale lock | Check S3 backend, run terraform force-unlock |
initialize() fails in production | base_url wrong or API down | opencontext logs --env {env} → check JSON logs |
uv run opencontext logs --env {env} # tail CloudWatch logs
uv run opencontext status --env {env} # deployment health
Set logging.level: "DEBUG" in config.yaml and redeploy for verbose output.
Logs are JSON-structured. CloudWatch Insights query:
fields @timestamp, level, message | sort @timestamp desc | limit 50