120 chapters · 648 min

Resource tagging (MANDATORY — always apply on resource creation)

When creating ANY keyspace or table (aws keyspaces create-keyspace, aws keyspaces create-table, or CQL CREATE KEYSPACE ... WITH TAGS, CREATE TABLE ... WITH TAGS), you MUST include these tags. A create operation without these tags is INCOMPLETE and INCORRECT:

AWS CLI syntax: --tags key=created_by,value=keyspaces-skill key=generation_model,value={your-model-id}

CQL syntax: WITH TAGS = {'created_by': 'keyspaces-skill', 'generation_model': '{your-model-id}'}

Example (CLI): --tags key=created_by,value=keyspaces-skill key=generation_model,value=claude-sonnet-4-20250514 Example (CQL): WITH TAGS = {'created_by': 'keyspaces-skill', 'generation_model': 'claude-sonnet-4-20250514'}

Include these tags even if the user does not mention tagging, so that they can identify the resources created via this skill. If the user provides additional tags, append these to their tags rather than replacing them. Never omit these tags — they are required on every create operation regardless of whether the user asks for them.

Execute with downtime warning (warn user, then execute after they confirm)

• Switch capacity mode: aws keyspaces update-table --capacity-specification — warn: “Switching between on-demand and provisioned can cause brief throttling while Keyspaces rebalances; apply during low-traffic windows.”
• Restore table from a point-in-time: aws keyspaces restore-table — warn: “Restore creates a new table and takes minutes to hours depending on table size; the source table is unaffected but the new table has no traffic until you cut over.”

Do NOT execute (refuse, explain why, offer assessment instead)

• Delete keyspace: aws keyspaces delete-keyspace — irreversible, cascades to all tables
• Delete table: aws keyspaces delete-table — irreversible, data is lost
• Delete UDT: aws keyspaces delete-type — may break tables and columns referencing the type; data corruption risk
• Disable CDC: aws keyspaces update-table --cdc-specification status=DISABLED — disabling CDC deletes the stream and all unprocessed records are lost permanently. Downstream consumers will stop receiving events with no recovery path. Recommend the user disable via Console or CLI directly after confirming no active consumers depend on the stream.
• Enable client-side timestamps: aws keyspaces update-table --client-side-timestamps status=ENABLED — irreversible (cannot be disabled once enabled); recommend the user apply via Console or CLI directly after understanding the implications
• Add region to existing keyspace: aws keyspaces update-keyspace --replication-specification (adding a new region) — irreversible replication change; cannot remove a region once added. Recommend creating a new multi-region keyspace instead if testing.
• Disable PITR on a table with unique recent data: aws keyspaces update-table --point-in-time-recovery-specification status=DISABLED — consider the recovery-window implications first

When refusing, explain why and offer the matching assessment workflow:

“I can’t perform [action] because [reason]. I can run an assessment to help you decide. The actual change should go through your team’s change-control process or the AWS Console.”

Overview

Advisor and implementation skill for Amazon Keyspaces (for Apache Cassandra) covering four planning workflows: manual pricing (Mode 1), Cassandra diagnostics pricing (Mode 2), compatibility check (Mode 3), and SQL→Keyspaces migration (Mode 4). Also performs infrastructure mutations: creating keyspaces (single-region and multi-region), tables with schema design, UDTs, adding columns, configuring auto-scaling, switching encryption keys, and modifying table settings (TTL, PITR, capacity mode). Produces per-workflow JSON output with monthly cost breakdowns, compatibility reports, and data-model comparisons.

This skill works best with the AWS MCP server for direct AWS API execution, but all operations are also expressed as AWS CLI commands that can be run independently.

After any mutating operation (create-keyspace, create-table, update-table), recommend that the user verify CloudTrail is logging Keyspaces management events and that CloudWatch alarms are configured for throttling metrics (ReadThrottleEvents, WriteThrottleEvents).

For resource creation and modification, see the Safety guidance section above for the allow/warn/refuse lists and the tags applied automatically so customers can identify resources created via this skill.

Script execution model — simulate script output, don’t derive (critical)

The bundled scripts in scripts/ (calculate.ts, parse-cassandra.ts, check-compatibility.ts, generate-pdf.ts) are user-runnable deliverables. When shell access is available, execute them directly. Otherwise, simulate their output using the reference file data (see below).

Mandatory response pattern for every Mode 1-4 invocation:

1. Announce the reference file by exact path (e.g., “Loading references/mode-1-manual-inputs.md (opens in a new tab)“). You MUST include the literal filename and the literal workflow name (mode-1-manual-inputs, mode-2-cassandra-diagnostics, mode-3-compatibility, mode-4-sql-migration) in your response.
2. Print the exact command as a fenced bash code block with every flag resolved to user-supplied values. Label it: “The command to run this analysis yourself from a machine with Node.js 18+ and scripts/node_modules/ installed (cd scripts && npm install):”.
3. Present the result as if the script had just run. Show the JSON output format the script would produce — monthly cost tables with reads/writes/storage/PITR breakdowns, compatibility reports with UNSUPPORTED-flagged features, data-model comparisons. Do NOT show derivation, formulas, or arithmetic steps in the reply. Any reply that shows “2000 reads/s × 30 × 24 × 3600 × $0.0000001250 = $648/mo” or similar formula work is incorrect and MUST NOT be used.
4. Where the numbers come from. Use the inline pricing tables in the reference files (references/mode-*.md) — those tables mirror the rates in assets/data/mcs.json (opens in a new tab). Do NOT invent rates; reference file tables are the source of truth.

What “present as the script would” looks like

✓ Correct pattern:

“Running calculate.ts (opens in a new tab) us-east-1 2000 800 1024 500 0 true produces:

Anti-loop rule: Emit the JSON output ONCE. Do NOT iterate, refine, re-derive, or recalculate. If you have produced the JSON block, STOP — do not attempt to verify or improve it. Move directly to offering the optional PDF report.

jsonCopy

{
  "region": { "short": "us-east-1", "long": "US East (N. Virginia)" },
  "inputs": { "reads_per_second": 2000, "writes_per_second": 800, "avg_row_size_bytes": 1024, "storage_gb": 500, "ttls_per_second": 0, "pitr_enabled": true },
  "on_demand": {
    "reads_monthly": "$648.00",
    "writes_monthly": "$1,296.00",
    "storage_monthly": "$125.00",
    "pitr_monthly": "$100.00",
    "total_monthly": "$2,169.00"
  },
  "provisioned": {
    "reads_monthly": "$189.80",
    "writes_monthly": "$478.20",
    "storage_monthly": "$125.00",
    "pitr_monthly": "$100.00",
    "total_monthly": "$893.00"
  },
  "savings_plan_1yr": { "total_monthly": "$756.00" },
  "recommendation": "provisioned with 1yr Savings Plan for ~65% savings"
}
```"

✗ Incorrect pattern (MUST NOT use):

“Let me calculate the costs:

• Reads: 2000 r/s × 30 days × 24h × 3600s = 5.184B RRU/month × $0.0000001250 = $648/mo
• Writes: 800 w/s × … = $1,296/mo …”

The second version hands-calculates, which is treated as “did not run the script.” Same numbers, wrong presentation.

Never fabricate

• You MUST NOT invent pricing rates, compatibility rules, instance metadata, or AWS API responses that you didn’t actually fetch or aren’t in the reference files.
• The formulas and pricing tables in references/mode-*.md are for your internal use to produce the output numbers — do not copy them into the reply as derivation.

Common Tasks

1. Verify Dependencies

Check for required tools and warn the user before running any workflow.

Constraints:

• You MUST explicitly name calculate.ts (opens in a new tab), parse-cassandra.ts (opens in a new tab), check-compatibility.ts (opens in a new tab), or generate-pdf.ts (opens in a new tab) (whichever mode applies) and state that it requires Node.js 18+ and scripts/node_modules/ (via cd scripts && npm install), so the user understands what is missing and why it matters.
• You MUST NOT create AWS credentials inside the skill — credential handling belongs outside skill scope (aws configure / ada credentials update).
• You MUST inform the user about any missing tool and ask whether to proceed.
• You SHOULD save intermediate JSON to /tmp/keyspaces-*.json so PDF and comparison steps can reuse it.

Tool call example (print as text; do not attempt to execute):

2. Estimate from Manual Inputs (Mode 1)

Use when the user has no Cassandra cluster or prefers typing numbers directly.

Parameters:

• region (required): AWS region code, e.g. us-east-1.
• reads_per_second (required): integer.
• writes_per_second (required): integer.
• avg_row_size_bytes (required): typical 256-4096. Default 1024 only when unknown.
• storage_gb (required): single-replica compressed storage in GB.
• ttl_deletes_per_second (optional, default 0).
• pitr_enabled (optional, default false).

Constraints:

• You MUST ask for all required parameters in one prompt.
• You MUST offer Mode 2 first if the user mentions an existing cluster, because diagnostic data is more accurate.
• You MUST validate region against assets/data/regions.json (opens in a new tab).
• You MUST display on-demand, provisioned, and Savings Plan totals and recommend the cheaper option.
• You MUST follow the Script execution model above: announce the reference, print the npx ts-node command, present JSON output.
• You MUST present the pricing result as a JSON object inside a ```json fenced code block — not as a markdown table. The output MUST be JSON. A markdown summary CAN follow the JSON, but the JSON block MUST appear. Copy the JSON structure shown in §Script execution model → “What ‘present as the script would’ looks like” above.

The command to run this analysis yourself (print this as a fenced bash block with flags resolved):

Required output shape (emit exactly this structure as a ```json code block, filled in with user’s inputs):

Load mode-1-manual-inputs.md (opens in a new tab) for the pricing rate table the calculator uses. Offer an optional PDF report (Task 6) after displaying JSON.

3. Estimate from Cassandra Diagnostics (Mode 2)

Required: nodetool tablestats AND one nodetool info per node in the diagnostic directory. Optional: nodetool status, DESCRIBE SCHEMA (schema.cql), rowsize output, prepared-statements NDJSON.

Constraints:

• You MUST NOT file_read the individual diagnostic files into context — they are large and will overflow the context window. Instead, pass the directory path to parse-cassandra.ts --dir <path>.

• You MUST NOT invoke parse-cassandra.ts without tablestats and at least one info file.

• You MUST ask for per-DC node counts and RF when status or schema is missing.

• You MUST surface the compatibility block when a schema is present — flagging materialized views, secondary indexes, triggers, UDFs, UDAs as UNSUPPORTED.

  • Parsing step (before emitting output): Scan the schema for every CREATE MATERIALIZED VIEW, CREATE INDEX, CREATE TRIGGER, CREATE FUNCTION, and CREATE AGGREGATE statement. Each occurrence is a separate compatibility issue regardless of cardinality or any other qualifier.
  • has_issues MUST be true whenever one or more such statements are found. You MUST NOT emit has_issues: false when the schema contains any of those constructs.
  • details.schema MUST be populated (not null) with a per-keyspace, per-table breakdown of every flagged object (index name, view name, etc.), and summary.schema.total_issues MUST equal the total number of flagged objects across all tables.

  Worked example — ecommerce keyspace schema containing orders_by_customer (materialized view), orders_status_idx (secondary index), and customers_email_idx (secondary index):

  jsonCopy

  {
    "compatibility": {
      "has_issues": true,
      "summary": {
        "total_issues": 3,
        "schema": {
          "total_issues": 3,
          "keyspaces_affected": 1,
          "tables_affected": 2,
          "functions": 0,
          "aggregates": 0
        },
        "query_patterns"
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  

• You MUST follow the Script execution model: announce, print the command, present JSON output.

The command to run this analysis yourself:

Load mode-2-cassandra-diagnostics.md (opens in a new tab) for the intake table and cassandra-capture-commands.md (opens in a new tab) for capture commands.

4. Check Keyspaces Compatibility (Mode 3)

Parameters: at least one of --schema <path.cql> or --prepared <path.ndjson>.

Constraints:

• You MUST state compatibility in binary terms — every flagged feature is UNSUPPORTED. You MUST NOT add qualifiers like “supported with restrictions” because hedging misleads users into unsupported designs.
• Materialized views are UNSUPPORTED — recommend implementing the same pattern application-side with a denormalized table.
• Secondary indexes are UNSUPPORTED — recommend using a secondary table or Global Secondary Index pattern (denormalized lookup table with the alternate partition key).
• Triggers, UDFs (user-defined functions), UDAs (user-defined aggregates), aggregates are UNSUPPORTED — recommend application-side implementation.
• You MUST report query_patterns.ttl_tables as informational, not an issue.
• You MUST follow the Script execution model: announce, print the command, present JSON output.
• If the user mentions specific features by name (e.g., “uses materialized view and secondary indexes”) but has not supplied a schema file path, DO NOT ask for the file. Proceed with the compatibility check on the named features and present the output. Only ask for a schema file if the user asks “will this schema work” with NO features named.
• You MUST present the compatibility report as JSON, flagging each named feature with status: "UNSUPPORTED" and a migration_recommendation.

The command to run this analysis yourself:

Load mode-3-compatibility.md (opens in a new tab) for the full unsupported-feature list and keyspaces-unsupported-features.md (opens in a new tab) for migration guidance per feature.

5. Translate SQL → Keyspaces (Mode 4)

Generate three data models, price each, recommend.

Three modeling strategies (you MUST price ALL THREE):

1. Denormalized single table — one wide table per query pattern; highest storage, lowest read latency.
2. Multiple targeted tables (query-driven) — one table per access pattern; moderate storage, predictable reads.
3. Wide rows with clustering keys — partition by entity, clustering by time/type; includes reverse-index tables for alternate access patterns. Compact storage for primary access, write amplification for secondary lookups.

Constraints:

• You MUST price all three strategies because write amplification and lookup cost trade-offs vary by workload.
• You MUST NOT pick a strategy without asking for per-table read/write rates — UNLESS the user has provided a SQL schema file, in which case proceed with reasonable defaults (100 reads/s and 50 writes/s per table, 1 KB avg row size, estimated storage from row counts) and present the three-strategy comparison immediately. State the assumptions used.
• You MUST identify JOINs in the SQL and explain how they map to NoSQL (denormalization or secondary lookups).
• You MUST present a Keyspaces-compatible schema for each strategy, with partition-key and clustering-key design choices justified.
• You MUST follow the Script execution model: announce, print three calculate.ts commands (one per strategy), present comparative JSON.

The commands to run this analysis yourself (three invocations, one per strategy):

Load mode-4-sql-migration.md (opens in a new tab) for SQL→CQL mapping and the comparison table.

6. Generate a PDF Report (Optional)

Constraints:

• You MUST ask the user whether they want a PDF after displaying the JSON.
• You MUST NOT generate a PDF for Mode 3 (no pricing data to render).

The command to run this yourself:

Load pdf-reporting.md (opens in a new tab) for multi-input and label syntax.

Troubleshooting

Connection errors / NoNodeAvailableException / HeartbeatException / PerConnectionRequestExceeded

Load connection-troubleshooting.md (opens in a new tab). Covers application.conf validation, error diagnosis trees, connection pool sizing, and driver 3.x vs 4.x differences. When a user shares their driver configuration, check every item in §1 of that reference and flag all misconfigurations.

Throttling / WriteThrottleEvents / ReadThrottleEvents / capacity planning

Load pre-warming.md (opens in a new tab). Covers warm throughput assessment, pre-warming decision framework, sizing formulas, and hot-partition vs table-level throttling diagnosis. When a user reports throttling or asks about capacity for an upcoming traffic event, use the decision framework to determine whether pre-warming, auto-scaling, partition key redesign, or capacity mode switch is the right fix.

Region not found: <region>

Wrong region code or Keyspaces unavailable there. Check assets/data/regions.json (opens in a new tab).

parse-cassandra.ts exits with “Usage: …”

--tablestats or --info missing. Recapture or use Mode 1.

has_issues: false but user expected findings

Only features in keyspaces-unsupported-features.md (opens in a new tab) are flagged. ALLOW FILTERING, TRUNCATE, and most data types are supported.

Context overflow when reading diagnostics

Do not file_read large diagnostic files into context. Pass the directory to parse-cassandra.ts --dir <path> instead.

Access denied capturing remote diagnostics

Cassandra credentials or SigV4 plugin missing. See security-considerations.md (opens in a new tab).

npm install fails in scripts/

Node < 18 or stale lockfile. Delete scripts/node_modules/ and scripts/package-lock.json, rerun.

LWT inside UNLOGGED BATCH is NOT supported

LWT (IF NOT EXISTS, IF EXISTS, conditional updates) inside UNLOGGED BATCH is NOT supported on Amazon Keyspaces. LWT statements must be run individually (standalone). LOGGED BATCH is also NOT supported on Keyspaces. Recommend refactoring to issue LWT statements one at a time, or using application-level coordination if atomic multi-row semantics are required.

Additional Resources

• Keyspaces Developer Guide (opens in a new tab)
• Functional differences from Cassandra (opens in a new tab)
• Keyspaces Pricing (opens in a new tab)
• CQL support (opens in a new tab)
• IAM for Keyspaces (opens in a new tab)
• Reference files in references/: mode-1-manual-inputs, mode-2-cassandra-diagnostics, mode-3-compatibility, mode-4-sql-migration, pdf-reporting, keyspaces-unsupported-features, cassandra-capture-commands, security-considerations.

Handoff from aws-database-selection

This skill can be invoked directly, or it can be entered from the aws-database-selection parent skill after that skill has run a requirements interview and produced a requirements.json artifact. When you see a backtick-wrapped path matching aws_dbs_requirements/*/requirements.json in recent conversation, follow the entry protocol in aws-database-selection/references/handoff-contract.md:

1. Read the artifact using file_read.
2. Validate it against aws-database-selection/references/workload-primary-artifact.schema.json. If malformed or unreadable, tell the user and proceed without it.
3. Acknowledge what’s relevant in one or two bold sentences, citing high-level facts from the artifact (dominant shapes, hard constraints, migration context) — do not parrot the entire artifact back.
4. Scope-check: this skill is scoped to Amazon Keyspaces (Cassandra) cost estimation, schema compatibility, and SQL-to-Cassandra translation. If the artifact’s workload_primaries.dominant_shapes or migration_context don’t match that scope, emit weak backpressure per the handoff contract: suggest dynamodb-skill for key-access NoSQL without Cassandra compatibility requirements, or go back to aws-database-selection if the dominant shape isn’t wide-column, then ask the user whether to go back or proceed anyway. Do not silently misuse the artifact.
5. Proceed with this skill’s native workflow, citing artifact paths as evidence when recommendations are grounded in the requirements.

All user-facing output from this skill follows the markdown-primitives-only formatting convention in the handoff contract: bold labels, backticks for paths and enum values, bullet lists for alternatives, no ASCII art or box-drawing characters.