Countly SDK Skill

An Agent Skill that teaches AI coding agents how to integrate Countly product analytics with a proper data collection model — so the resulting dashboard can actually answer product questions (funnels, retention, cohorts, drill), instead of collecting data that silently hits server limits or splits users.

What's inside:

• skills/countly-integration/SKILL.md — the core guide: design the data model backwards from dashboard questions, identity/login/logout strategy and merge semantics, event & segmentation cardinality rules, user property budget, view-name normalization, crashes, consent, server limits, and a verification checklist.
• references/sdk-methods.md — exact method names per SDK (Web/JS, Android, iOS, Flutter, React Native).
• references/http-api.md — the raw /i HTTP ingestion API for server-side integrations.

The guidance is verified against the Countly server source and the official docs at support.countly.com.

Install

Claude Code (recommended — automatic triggering)

As a plugin, kept up to date:

/plugin marketplace add Countly/countly-sdk-skill
/plugin install countly-integration@countly

Or copy the skill folder directly:

git clone https://github.com/Countly/countly-sdk-skill
mkdir -p ~/.claude/skills
cp -R countly-sdk-skill/skills/countly-integration ~/.claude/skills/

(Use a project's .claude/skills/ instead of ~/.claude/skills/ to share it with your team via the repo.)

Claude.ai / Claude Desktop

Upload dist/countly-integration.skill under Settings → Capabilities → Skills.

Cursor

cp dist/cursor/countly.mdc your-project/.cursor/rules/

Codex & other AGENTS.md-aware tools

Append dist/AGENTS.md to your project's AGENTS.md.

GitHub Copilot

Append dist/copilot-instructions.md to .github/copilot-instructions.md.

Any other LLM or agent

dist/llms-full.txt is the entire skill flattened into one self-contained markdown file. Paste it into your system prompt / project knowledge, or tell a file-reading agent: "Read llms-full.txt and follow it before integrating Countly."

Repo layout

skills/countly-integration/   source of truth (Agent Skills format)
dist/                         generated formats — do not edit by hand
scripts/build.py              regenerates dist/ from the skill
evals/evals.json              test prompts + assertions used to benchmark the skill

Contributing

Edit only skills/countly-integration/, then regenerate the distribution formats:

python3 scripts/build.py

CI regenerates dist/ on every push to main and fails if it's out of sync, so the formats can never drift from the skill.

The skill is benchmarked with the prompts and assertions in evals/evals.json; if you change guidance (limits, identity semantics, API names), keep the assertions in step.

License

MIT