Labsco
Labsco/Repos/streamlit/agent-skills
REPO PACKAGE

streamlit agent-skills

COMMUNITY
streamlit · publisher213 repository starsApache-2.0 · Freegithub.com/streamlit/agent-skills
1skill
0ready to use

Starting with Streamlit 1.57, the actual skill content — dashboards, theming, custom components, session state, performance — moved out of this repository and into the streamlit package. The one skill left here, developing-with-streamlit, detects your active Python interpreter (virtualenv, conda, pipenv, poetry, pdm, uv, or system), locates the installed Streamlit path, and points the agent at <streamlit_path>/.agents/skills/; on Streamlit versions older than 1.57 it falls back to scraping Streamlit's own online docs instead.

This is for anyone building with Streamlit 1.57 or newer who wants one install to keep working across every project and every future Streamlit upgrade, since the meta-skill resolves against whichever interpreter is active rather than pinning a version. On an older Streamlit, or a project with no discoverable interpreter, the fallback is generic doc-scraping rather than the version-matched reference material the newer path provides.

READ THE FULL ANALYSIS

Why the numbers look like a collapse. Nineteen of the twenty skills this repository has ever carried are now dead — not abandonment, but the architecture change the README describes: the real per-feature skills were relocated into the streamlit pip package as of 1.57, and this repository was cut down to the one discovery skill that finds them.

Getting it installed is not one command. streamlit skills only exists from Streamlit 1.58 onward; on 1.57 the README routes you through several different manual paths depending on your agent (a raw git clone and copy for Claude Code, gh skill install for Copilot and Cursor, gemini skills install for Gemini CLI, or the cross-agent npx skills add installer), and Snowflake Cortex Code is the only client that ships this by default.

213Stars on the GitHub repository, at last check.