GlassFlow Agent Skills

Skills for driving GlassFlow ETL from any AI coding assistant that reads SKILL.md files, including Claude Code, Cursor, and OpenAI Codex. Each skill is an instruction set the assistant loads when you describe a matching task, so you can create and operate GlassFlow pipelines from natural language.

Install

Pick your AI coding tool.

Claude Code

/plugin marketplace add glassflow/agent-skills
/plugin install glassflow-agent-skills@glassflow-agent-skills

Update later with /plugin marketplace update glassflow-agent-skills.

Cursor

git clone https://github.com/glassflow/agent-skills.git /tmp/glassflow-agent-skills
mkdir -p .cursor/skills
cp -r /tmp/glassflow-agent-skills/skills/* .cursor/skills/

Reload Cursor so it picks up the new files. Cursor reads SKILL.md frontmatter natively.

OpenAI Codex

git clone https://github.com/glassflow/agent-skills.git /tmp/glassflow-agent-skills
mkdir -p .codex/skills
cp -r /tmp/glassflow-agent-skills/skills/* .codex/skills/

For global availability across all your projects, copy into ~/.codex/skills/ instead.

Other tools

The skills are plain SKILL.md files. Clone the repo and copy skills/* into the path your tool expects:

Tool	Path
Claude Code (manual, no plugin)	.claude/skills/
Cursor	.cursor/skills/
OpenAI Codex	.codex/skills/ (or ~/.codex/skills/ for global)
OpenCode	.opencode/skills/
GitHub Copilot / VS Code	.github/skills/

A community CLI automates this across tools:

npx skills add glassflow/agent-skills

Usage

Trigger a skill by describing the task, for example:

"Create a Kafka to ClickHouse pipeline for the orders topic with deduplication on order_id over 1 hour."

Available skills

Skill	Description
create-pipeline	Create a GlassFlow ETL pipeline from natural language. Walks through source, optional dedup/filter/transform stages, and ClickHouse sink mapping, then submits the v3 config to the GlassFlow API and waits for the pipeline to reach Running.
debug-pipeline	Localize a pipeline failure or missing-events problem. Starts with component logs (via kubectl or the OTel Collector's log backend), then cross-checks with metrics for silent failures, backpressure, and DLQ rate.
tune-pipeline	Resolve sustained backpressure or sink-lag by reading GlassFlow's emitted metrics to confirm the bottleneck, proposing targeted changes to replicas, batch sizes, stream limits, or dedup storage, applying them via the appropriate API endpoint, and verifying by re-reading the same metrics.

More skills are added as we surface common workflows. See the docs for the canonical list and walkthrough.

Contributing

This repository carries skills that are useful to GlassFlow users in general. Internal team workflows live in a private repo and are not published here.

To propose a new skill, open an issue describing the user-visible task, the inputs the skill needs, and the steps the agent should take. Or open a PR with a draft SKILL.md under skills/<skill-name>/ following the structure in skills/create-pipeline/.

License

Apache-2.0. See LICENSE.