Skip to content

Agent skill

The using-pydynox skill helps coding assistants build applications with pydynox. It explains how to choose an operation, handle its results, and test the application's behavior with async and sync APIs.

Task What the skill covers
Pagination Streaming, "load more", limit vs page_size, cursor boundaries, empty filtered pages, and exhaustion
Queries and indexes Sort-key ranges, filters, descending order, GSI/LSI, projections, scans, and counts
Writes Create-only conditions, atomic updates, batch operations, transaction rollback, and retry decisions
Models Attributes, aliases, Decimal, key templates, TTL, smart save, hooks, and optimistic locking
Integrations Pydantic, dataclasses, S3, KMS, compression, and feature-specific guide links
Operations Credentials, timeouts, consistency, throttling, concurrency, and metrics
Testing MemoryBackend limits and executable LocalStack recipes

Install

From your application project, run:

npx skills add ferrumio/pydynox --skill using-pydynox

The installer lets you choose which coding assistant receives the skill. It requires Node.js/npm. This installs agent instructions in your project; install the Python library in your application environment as usual:

pip install pydynox

You can also copy the complete skills/using-pydynox directory into the skill directory supported by your coding assistant. Keep its references/ and examples/ directories together.

Use it

Ask your coding assistant to use using-pydynox, for example:

Use using-pydynox to add an Order model, query orders by customer, and test the application without AWS.

Or ask for a specific application behavior:

Use using-pydynox to implement "load more" for a customer's paid orders. Return a continuation cursor, keep descending order, and test that no orders are skipped when some DynamoDB pages contain no matches.

The skill checks the application's installed version and routes the assistant to a focused reference and runnable Python recipes. Pagination is also explained in the main SKILL.md, including the cursor problem when iteration stops with unread items in a fetched page.

The basic example runs on pydynox 1.8.0 with MemoryBackend, or with LocalStack for pagination and transactions. The additional query recipes use LocalStack for GSI, projections, and scans. Empty filtered-page recipes require the fix merged in PR #565 after 1.8.0. The skill's testing reference contains the commands and version notes.

The skill follows the Agent Skills format. Its source lives beside the library, and the Python test suites exercise its examples. Contributors working on pydynox itself should start with the repository's AGENTS.md.