Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Database Architect · dokhacgiakhoa | DecimalAI

dokhacgiakhoa/database-architect

NewSkillScoreMaintained

Expert database architect specializing in data layer design from scratch, technology selection, schema modeling, and scalable database architectures. Masters SQL/NoSQL/TimeSeries database selection, normalization strategies, migration planning, and performance-first design. Handles both greenfield architectures and re-architecture of existing systems. Use PROACTIVELY for database architecture, technology selection, or data modeling decisions.

Databy dokhacgiakhoa

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

Data Engineer pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/dokhacgiakhoa-database-architect/SKILL.md

decimalai skills pull dokhacgiakhoa-database-architect --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/dokhacgiakhoa-database-architect/SKILL.md https://app.decimal.ai/s/dokhacgiakhoa-database-architect/SKILL.md

Raw file: /s/dokhacgiakhoa-database-architect/SKILL.md ↗ · machine-readable summary at /s/dokhacgiakhoa-database-architect.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 3h ago

Yes — +5% lift · at higher cost+38% tokens0% turns86% of whole cases pass with skill (19/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-02	✗→✓	▲ Improved	32%	0%
case-01	✓→✓	＝ Same ✓	60%	0%
case-03	✓→✓	＝ Same ✓	54%	0%
case-04	✓→✓	＝ Same ✓	6%	0%
case-05	✓→✓	＝ Same ✓	51%	0%

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/10/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

dokhacgiakhoa / agent-skills-4-vibe-coding-cli

.agent/skills/database-architect/SKILL.md

View on GitHub ↗

License NOASSERTIONImported 8/10/2026Version v1

Skill contents

SKILL.md

You are a database architect specializing in designing scalable, performant, and maintainable data layers from the ground up.

Use this skill when

Selecting database technologies or storage patterns
Designing schemas, partitions, or replication strategies
Planning migrations or re-architecting data layers

Do not use this skill when

You only need query tuning
You need application-level feature design only
You cannot modify the data model or infrastructure

Instructions

Capture data domain, access patterns, and scale targets.
Choose the database model and architecture pattern.
Design schemas, indexes, and lifecycle policies.
Plan migration, backup, and rollout strategies.

Safety

Avoid destructive changes without backups and rollbacks.
Validate migration plans in staging before production.

Purpose

Expert database architect with comprehensive knowledge of data modeling, technology selection, and scalable database design. Masters both greenfield architecture and re-architecture of existing systems. Specializes in choosing the right database technology, designing optimal schemas, planning migrations, and building performance-first data architectures that scale with application growth.

Core Philosophy

Design the data layer right from the start to avoid costly rework. Focus on choosing the right technology, modeling data correctly, and planning for scale from day one. Build architectures that are both performant today and adaptable for tomorrow's requirements.

Capabilities

🧠 Knowledge Modules (Fractal Skills)

1. Technology Selection & Evaluation

2. Data Modeling & Schema Design

3. Normalization vs Denormalization

4. Indexing Strategy & Design

5. Query Design & Optimization

6. Caching Architecture

7. Scalability & Performance Design

8. Migration Planning & Strategy

9. Transaction Design & Consistency

10. Security & Compliance

11. Cloud Database Architecture

12. ORM & Framework Integration

13. Monitoring & Observability

14. Disaster Recovery & High Availability

Want to see if dokhacgiakhoa/database-architect would activate on your phrasing? Try it before you install — no LLM call, no signup.

Latest benchmark · publisher runVerified

8/29/2026, 07:14 UTC · 22 cases

Lift

+5%

86% vs 82%

Agent turns

1.0 vs 1.0

Tokens

+38%

3,555.82 vs 2,575.68

fail → pass · 1pass → pass · 18fail → fail · 3

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-01 We are designing a core banking ledger for a new fintech product requiring strict ACID transactions, multi-table consistency, and complex financial audits. Team members suggest using MongoDB for rapid early iteration. What database paradigm should be selected for the primary storage model?	pass→pass	18,465	29,968	+62%	1	1	0%	2,444	3,901	+60%	0	0	—
case-02 An industrial IoT platform ingests 50,000 sensor telemetry readings per second. Queries focus heavily on time-window aggregations (e.g., average temperature per hour). The team wants to store this in a standard MySQL database using standard InnoDB tables. How should the storage layer be designed?	fail→pass	24,421	32,667	+34%	1	1	0%	4,149	5,464	+32%	0	0	—
case-03 A web application experiences high latency because user session validation reads from PostgreSQL on every HTTP request. Engineers propose creating a `sessions` table in Postgres with an unindexed `last_active` timestamp. How should user session storage be architected?	pass→pass	18,516	20,582	+11%	1	1	0%	2,622	4,045	+54%	0	0	—
case-04 In a 500GB PostgreSQL production table `orders`, we need to add a mandatory non-nullable column `status_code` with default `'pending'`. Running a direct `ALTER TABLE orders ADD COLUMN status_code VARCHAR NOT NULL DEFAULT 'pending';` will lock the table for extended periods. What migration strategy should be executed?	pass→pass	18,922	16,448	-13%	1	1	0%	3,255	3,464	+6%	0	0	—
case-05 A B2B SaaS platform is partitioning its database across multiple shards to support massive write growth. The lead engineer proposes sharding tables using auto-incrementing sequential `id` values to balance write distribution. Which sharding key strategy should be chosen instead?	pass→pass	11,618	13,114	+13%	1	1	0%	1,891	2,854	+51%	0	0	—
case-06 Users complain that after updating their profile picture, refreshing the page still shows the old picture due to PostgreSQL asynchronous replication lag to read replicas. Developers propose increasing replica CPU size. How should read consistency be guaranteed immediately after updates?	pass→pass	16,495	16,285	-1%	1	1	0%	2,348	3,348	+43%	0	0	—
case-07 An audit log table in PostgreSQL receives 10 million rows daily and retains data for 90 days. Deleting expired records via `DELETE FROM audit_logs WHERE created_at < NOW() - INTERVAL '90 days';` causes massive table bloat and high I/O spikes. What database table design should replace this?	pass→pass	14,750	24,021	+63%	1	1	0%	2,651	4,238	+60%	0	0	—
case-08 A frequent query executes `SELECT * FROM tickets WHERE organization_id = $1 AND status = 'open' ORDER BY created_at DESC LIMIT 20;`. Engineers created three separate single-column indexes on `organization_id`, `status`, and `created_at`. What index structure should be created instead?	pass→pass	11,858	10,989	-7%	1	1	0%	1,796	2,540	+41%	0	0	—
case-09 A professional networking service needs to calculate 2nd and 3rd degree connection paths between 5 million users. Developers are writing recursive SQL common table expressions (CTEs) over a relational `connections` JOIN table. What specialized data model handles this access pattern efficiently?	pass→pass	12,993	15,754	+21%	1	1	0%	2,059	3,280	+59%	0	0	—
case-10 An e-commerce company's analytics dashboard runs complex analytical aggregations directly against the production PostgreSQL primary OLTP database, degrading checkout performance. Developers suggest adding horizontal read replicas. What architectural pattern should be implemented instead?	pass→pass	12,710	17,787	+40%	1	1	0%	2,055	3,007	+46%	0	0	—
case-11 A mission-critical financial database in AWS us-east-1 requires a disaster recovery strategy with an RPO under 1 minute for regional outages. The ops team suggests taking nightly EBS volume snapshots and copying them cross-region. How should the disaster recovery data stream be configured?	pass→pass	14,499	20,083	+39%	1	1	0%	2,403	3,844	+60%	0	0	—
case-12 An API endpoint using SQLAlchemy loads 100 `User` objects, then iterates over each user in a loop accessing `user.orders`, causing 101 SQL queries. How should the data fetching strategy be corrected at the ORM layer?	pass→pass	9,600	14,077	+47%	1	1	0%	1,888	2,571	+36%	0	0	—
case-13 A system deploys 500 concurrent AWS Lambda instances connecting directly to a PostgreSQL database, causing 'too many connections' errors during traffic spikes. The team suggests raising `max_connections` to 10,000 in `postgresql.conf`. What middleware component should be deployed?	pass→pass	9,921	11,251	+13%	1	1	0%	1,477	2,208	+49%	0	0	—
case-14 To protect against corrupting database migrations, a database administrator sets up `pg_dump` logical backups every 6 hours. However, this risks losing up to 6 hours of user transaction data upon restore. What backup regime is required for point-in-time recovery?	pass→pass	12,401	13,527	+9%	1	1	0%	2,147	3,141	+46%	0	0	—
case-15 A primary transactional PostgreSQL database is 4TB in size, with 80% of data being historical records untouched for over 2 years. Storage costs are accelerating. How should this cold historical data be managed?	pass→pass	22,761	25,359	+11%	1	1	0%	3,161	4,348	+38%	0	0	—
case-16 A clickstream event reporting engine processes 100 million event rows daily. Ad-hoc query dashboards execute `SUM` and `COUNT GROUP BY` across 5 columns out of 50 total attributes. Standard MySQL InnoDB tables exhibit slow scan times. What table storage layout solves this?	pass→pass	14,445	16,345	+13%	1	1	0%	2,297	3,698	+61%	0	0	—
case-17 In an inventory booking system where item contention is infrequent, developers enforce row-level `SELECT FOR UPDATE` pessimistic locks during browsing, causing database connection exhaustion. What concurrency control model should be implemented?	pass→pass	10,807	19,735	+83%	1	1	0%	1,779	3,469	+95%	0	0	—
case-18 A catalog containing 10 million SKUs uses SQL `LIKE '%keyword%'` queries across product titles and descriptions, leading to full table scans and database CPU exhaustion. What dedicated component should handle this query pattern?	pass→pass	12,022	12,359	+3%	1	1	0%	1,590	2,518	+58%	0	0	—
case-19 The operations team wants to measure replication health on a PostgreSQL secondary node by comparing row counts in key tables every hour. What specific system telemetry metric should be monitored continuously instead?	pass→pass	10,761	11,746	+9%	1	1	0%	1,639	2,583	+58%	0	0	—
case-20 Given a specific query `SELECT id, name FROM users WHERE lower(email) = 'test@example.com'`, provide the exact EXPLAIN ANALYZE execution plan breakdown and rewrite the query without altering table indexes or schema structures.	fail→fail	14,735	21,681	+47%	1	1	0%	2,700	3,840	+42%	0	0	—
case-21 Design the React UI component forms, client-side input validation states, and submit button handlers for our new user registration flow.	fail→fail	34,314	28,879	-16%	1	1	0%	7,291	5,594	-23%	0	0	—
case-22 Write a SQL query using window functions to calculate monthly active user retention on our existing fixed database, where we cannot change any indexes or table definitions.	fail→fail	20,317	17,942	-12%	1	1	0%	3,023	4,273	+41%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +5 percentage points is the difference between those two pass rates over the 22 comparable cases.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 81.8%
Pass rate with the skill: 86.4%
Cases attempted: 22
Cases counted toward lift: 22
Cases that regressed: 0
Skill version measured: v1
Run completed: 8/29/2026, 07:14 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-02

Prompt

An industrial IoT platform ingests 50,000 sensor telemetry readings per second. Queries focus heavily on time-window aggregations (e.g., average temperature per hour). The team wants to store this in a standard MySQL database using standard InnoDB tables. How should the storage layer be designed?

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signalsno signals yet

BenchmarkRun a verified benchmark with positive lift

Live evalNeeds ≥20 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

No signals yet — run a benchmark or collect evaluated traces to earn a score.

Verified on gemini-3.6-flash · re-tested today

Other measured skills in the registry, with their headline benchmark lift.