Requirements
- Target platform
- OpenClaw
- Install method
- Manual import
- Extraction
- Extract archive
- Prerequisites
- OpenClaw
- Primary doc
- SKILL.md
Choose and architect storage systems for applications with the right tradeoffs.
Choose and architect storage systems for applications with the right tradeoffs.
Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually.
I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete.
I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run.
Object storage (S3, R2, GCS) for immutable blobs: images, videos, backups, logs โ cheap, scales infinitely, but no partial updates Block storage (EBS, Persistent Disks) for databases and apps needing filesystem semantics โ faster, but tied to single instance Network file systems (NFS, EFS) when multiple instances need shared filesystem access โ convenient but latency and cost add up Default to object storage for user uploads โ block storage for database files only
SQL when you need joins, transactions, or complex queries โ fighting against NoSQL for relational data wastes months Document stores (MongoDB, Firestore) for nested/variable schemas where you always fetch the whole document Key-value (Redis, DynamoDB) for simple lookups by ID at massive scale โ not for complex queries Time-series databases (InfluxDB, TimescaleDB) for metrics with timestamp-based queries โ regular SQL struggles with retention policies Start with PostgreSQL unless you have a specific reason not to โ it handles JSON, full-text search, and scales further than most assume
Local disk for ephemeral data: temp files, build artifacts, caches โ assume it disappears on restart Cloud storage for anything that must survive instance termination โ never store user data only on local disk Local SSD for databases in production โ network-attached storage adds latency to every query Hybrid: local cache in front of cloud storage for frequently accessed files
Put CDN in front of static assets always โ origin requests are slower and more expensive Set long cache TTLs with versioned URLs (style.abc123.css) โ cache invalidation is slow and unreliable CDN for dynamic content only if latency matters more than freshness โ adds complexity for marginal gains Edge caching for API responses works but cache keys get tricky โ start simple, add only when needed
Never accept uploads directly to app server disk in production โ use presigned URLs to cloud storage Set file size limits at load balancer level, not just application โ prevents memory exhaustion attacks Generate unique keys for uploads (UUIDs) โ user-provided filenames cause collisions and path traversal risks Validate file types by content (magic bytes), not extension โ extensions are trivially spoofed
Keep compute and storage in same region โ cross-region data transfer adds latency and cost Replicate data to regions where users are, not where developers are Multi-region storage adds complexity โ single region with backups elsewhere usually sufficient Database read replicas in user regions for read-heavy workloads
Define retention policy before storing data โ "keep everything" becomes expensive and legally risky Automate deletion of temporary data โ manual cleanup never happens consistently Tiered storage for aging data: hot โ warm โ cold โ archive โ but check retrieval costs before archiving Separate storage for logs vs business data โ different retention, different compliance requirements
Egress fees dominate cloud storage costs โ calculate before choosing provider Many small files cost more than few large files โ batch small writes when possible Minimum storage duration on cold tiers โ early deletion still charges full period API request costs matter at scale โ millions of LIST operations add up
3-2-1 rule: 3 copies, 2 different media types, 1 offsite โ cloud counts as one location Test restores regularly โ untested backups are not backups Point-in-time recovery for databases โ daily snapshots lose a day of data Version important files โ deletion or corruption often discovered late
Code helpers, APIs, CLIs, browser automation, testing, and developer operations.
Largest current source with strong distribution and engagement signals.