Datacake Agent Skill

An Agent Skill that teaches Claude Code and other AI agents how to build tools, analytics, scripts and custom web or mobile frontends on the Datacake IoT platform using its GraphQL API (https://api.datacake.co/graphql/).

The skill bundles:

• skills/datacake/SKILL.md – the entry point: Datacake mental model, essential queries, decision guide, hard rules and workflows.
• skills/datacake/reference/ – platform concepts, API basics, device/measurement/semantics query references, mutations, a curated schema map, the full GraphQL schema (SDL), and playbooks for Next.js and Expo apps and analytics scripts.
• skills/datacake/scripts/ – dependency-free Python 3 helpers: dc.py (run GraphQL operations), discover.py (map a workspace: products, field identifiers, tags, semantics), history_to_csv.py (historical data to CSV), fetch_schema.py (refresh the bundled schema).

Install

Claude Code (personal, all projects)

git clone <this repo>~/datacake-claude-skill
mkdir -p ~/.claude/skills
ln -s ~/datacake-claude-skill/skills/datacake ~/.claude/skills/datacake # or copy the folder

Start claude, then ask anything about Datacake or type /datacake. Claude loads the skill automatically when the request mentions Datacake.

Claude Code (project only)

Copy or symlink skills/datacake to <repo>/.claude/skills/datacake and commit it.

Claude Code (plugin)

claude plugin install /path/to/datacake-claude-skill --scope local# or: claude --plugin-dir /path/to/datacake-claude-skill

The plugin exposes the skill as /datacake:datacake (and /datacake when the name is free).

claude.ai (web, desktop, Cowork)

Zip the skills/datacake folder and upload it under Settings > Capabilities > Skills. The frontmatter only uses fields from the Agent Skills specification, so the upload validates.

Other agents (Cursor, Codex, agentskills.io-compatible tools)

Copy skills/datacake into the agent's skills directory (for example .cursor/skills/datacake/ or .codex/skills/datacake/). Everything is plain Markdown plus Python; script paths are relative to the skill folder.

Token setup

Create a token in Datacake (Account Settings > API Token, or a scoped API user under Members > API Users) and expose it to the scripts:

export DATACAKE_TOKEN=... # preferred# or
mkdir -p ~/.datacake &&echo -n "...">~/.datacake/token && chmod 600 ~/.datacake/token
python3 skills/datacake/scripts/dc.py 'query { user { id email } }'
python3 skills/datacake/scripts/discover.py

Never commit tokens. The skill instructs agents to keep tokens server-side in any app they build.

Keeping the schema current

python3 skills/datacake/scripts/fetch_schema.py --check # what changed since the bundled schema
python3 skills/datacake/scripts/fetch_schema.py # rewrite reference/schema.graphql and the generated blocks in reference/schema-map.md

Development

python3 -m pip install --user graphql-core
python3 tools/validate_examples.py # every ```graphql block in the skill validates against the schema
DATACAKE_TOKEN=... python3 tools/smoke_test.py # runs the canonical queries against a real workspace (read-only)
claude plugin validate . --strict # frontmatter and manifest checks

Reference files are written for agents: terse, tables, one validated example per operation. Keep SKILL.md under 400 lines and put detail into reference/.

License

MIT. Datacake is a product of Datacake GmbH; this skill is documentation and tooling around its public API.