Amazon Aurora PostgreSQL

A modular toolkit for Aurora PostgreSQL organized as a registry of sub-skills. Each sub-skill handles one domain of Aurora PostgreSQL work. The router matches user intent to the right sub-skill, then loads only the references needed. (For Aurora MySQL, use the amazon-aurora-mysql skill.)

Operating procedure (follow in order)

1. Route — match the request to a sub-skill using the Trigger phrases column (match on meaning, not exact wording), then confirm with the When to route here column.
2. Load — file_read the matched sub-skill’s references/{id}-instructions.md and announce the path. Do not answer a matched sub-skill from general knowledge alone.
3. Analyze / advise — perform the sub-skill’s work; run a bundled script when the user supplies the inputs (see Scripts).
4. If a mutation is requested — classify against the Safety guardrails tier, confirm with the user, apply resource tags, then execute (MCP-preferred, CLI fallback).
5. Present results — tables with dollar/ACU figures and a recommendation label; no derivation or arithmetic steps.

Edge cases: if the request spans multiple sub-skills, run them in sequence (load each instructions.md in turn). If no sub-skill matches, answer directly from Aurora PostgreSQL knowledge. If a script or MCP/CLI call fails, show the error and suggest a fix before retrying. The numbered Global rules below are details that hang off these steps.

Sub-skill registry

Column semantics: Trigger phrases = the keyword index you match the request against (step 1). When to route here = the decision logic confirming the match. Next steps = sub-skills to offer the user as a natural follow-up after this one completes (not auto-chained); Reached from = sub-skills that typically route into this one. Next-steps/Reached-from are suggestions for guiding the user, never automatic execution.

Express vs Full configuration — decision matrix

When routing a create request (sub-skill create), pick the path with this matrix. Express is the default for Aurora PostgreSQL; route to Full configuration only if ANY “Full” trigger is present. Don’t present the choice to the user — decide, then state which path and why.

Notes: any single Full trigger disqualifies express — name every trigger you matched in the routing statement. Express clusters are still customizable after creation (e.g. a custom parameter group can be applied post-create), so a future need isn’t itself a reason to start with Full. Full depth on the flow lives in references/express-create-instructions.md and references/create-instructions.md — load those for the actual steps.

Global rules (apply to every sub-skill)

1. Execute, don’t just suggest. When the user requests an action and confirms, EXECUTE it rather than handing back a command to run. The AWS MCP server is the recommended execution path when available (sandboxed, IAM-authenticated, audit-logged) — prefer it. When MCP tools are not available (e.g. Claude Code, Cursor, or other non-MCP hosts), use the AWS CLI / SDK directly with the same aws rds ... operation. Only if execution is genuinely not possible in the current environment, present the complete CLI command for the user to run.

2. Confirmation before mutation. MUST confirm with the user before any create or modify operation. Do NOT execute without explicit confirmation (“yes”, “proceed”, “confirmed”, “go ahead”).

3. Resource tagging (always apply on resource creation). When creating any cluster or instance, ALWAYS include these tags: --tags Key=created_by,Value=aurora-skill Key=generation_model,Value={your-model-id} Use your model id if known; if you cannot reliably determine it, use Value=unknown — never let tagging block the create. Include these tags even if the user does not mention tagging. If the user provides additional tags, append these to their tags.

4. Safety guardrails.

   Tier 1 — Confirm (a yes/no confirmation is enough; no risk briefing required):

   • create-db-cluster, create-db-cluster --with-express-configuration
   • create-db-instance
   • modify-db-cluster --serverless-v2-scaling-configuration (ACU scaling)
   • modify-db-cluster --backup-retention-period
   • modify-db-cluster --deletion-protection / --no-deletion-protection
   • modify-db-cluster --enable-cloudwatch-logs-exports
   • modify-db-cluster --preferred-backup-window
   • modify-db-cluster --enable-http-endpoint (Data API)
   • add-tags-to-resource, remove-tags-from-resource

   Tier 2 — High-impact: state the specific risk, THEN confirm (spell out the impact before asking; do not call any API until the user confirms with that risk in front of them):

   • modify-db-cluster --storage-type — no downtime for most instance classes; requires restart for NVMe/Optimized Reads instances (r6gd, r6id, r8gd). Switching from Aurora Standard to Aurora I/O-Optimized is limited to once every 30 days; switching from Aurora I/O-Optimized back to Aurora Standard can be done at any time.

Scripts

Bundled scripts in scripts/ for offline analysis. MUST use these when the user provides the required inputs — do NOT hand-calculate. Each script documents its full flags/usage in its own --help and header docstring; read those on demand rather than relying only on the one-line usage below.

Script execution model: If a shell is available, execute the script directly and present the output. If no shell is available, print the exact command as a fenced bash code block with all flags resolved to user-supplied values, then present results computed inline from the reference file’s pricing tables. (Result-presentation format is governed by the Operating procedure / Global rules — no derivation steps.)

Troubleshooting

• AccessDenied: Attach AmazonRDSReadOnlyAccess + CloudWatchReadOnlyAccess for reads. For creates/modifies, use a custom policy scoped to rds:CreateDBCluster, rds:CreateDBInstance, rds:ModifyDBCluster, rds:ModifyDBInstance, rds:AddTagsToResource, and rds:Describe*. See Identity and access management for Amazon Aurora (opens in a new tab).
• ExpiredToken / credentials: Refresh your AWS credentials using whatever mechanism you use (e.g. re-run your SSO/aws sso login, ada credentials update, assume-role, or refresh the profile), then retry. Do not assume a specific credential tool.
• DBClusterNotFoundFault: Verify region and cluster ID.
• Throttling: Retry once, then narrow scope.

Additional Resources

• Aurora User Guide (opens in a new tab)
• Aurora pricing (opens in a new tab)
• Aurora serverless (opens in a new tab)
• Aurora PostgreSQL upgrades (opens in a new tab)

Handoff from aws-database-selection

This skill can be entered from aws-database-selection after it produces a requirements.json. When you see a path matching aws_dbs_requirements/*/requirements.json in conversation:

1. Read the artifact. Sanity-check it has the fields you’ll use — at minimum engine (or workload type), region, and the workload signals you route on (capacity/ACU hints, storage size, connectivity/VPC needs, version). If those are present and parseable, use them; if it’s missing them or won’t parse, proceed without it (don’t block on a formal schema).
2. Acknowledge relevant facts in 1-2 bold sentences.
3. Scope-check: if the artifact doesn’t match Aurora (e.g., key-access → DynamoDB, graph → Neptune, multi-region strong SQL → DSQL), suggest the right skill and ask whether to proceed anyway.
4. Continue with this skill’s sub-skill routing.