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/assets/brand/– Datacake logo files (wordmark black/white, icon mark, favicon, app icons) used as the default branding of generated frontends when the user supplies none; colour tokens and usage inreference/branding.md. Trademark notice inassets/brand/README.md.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).
git clone <this repo> ~/datacake-claude-skill
mkdir -p ~/.claude/skills
ln -s ~/datacake-claude-skill/skills/datacake ~/.claude/skills/datacake # or copy the folderStart claude, then ask anything about Datacake or type /datacake. Claude loads the skill automatically when the request mentions Datacake.
Copy or symlink skills/datacake to <repo>/.claude/skills/datacake and commit it.
claude plugin install /path/to/datacake-claude-skill --scope local # or: claude --plugin-dir /path/to/datacake-claude-skillThe plugin exposes the skill as /datacake:datacake (and /datacake when the name is free).
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.
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.
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 # workspaces; add --orgs for organizations
python3 skills/datacake/scripts/members.py list <workspace> # members, invites, API users (--csv for export)Admin tooling (organizations, members, invites, white label users, audit log) is covered in skills/datacake/reference/organizations-and-members.md; members.py invite|move|remove run the bulk operations as a dry run unless --execute is given.
Never commit tokens. The skill instructs agents to keep tokens server-side in any app they build.
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.mdpython3 -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)
python3 tools/check_assets.py # brand asset files present in skills/datacake/assets/brand/
claude plugin validate . --strict # frontmatter and manifest checksReference files are written for agents: terse, tables, one validated example per operation. Keep SKILL.md under 400 lines and put detail into reference/.
MIT. Datacake is a product of Datacake GmbH; this skill is documentation and tooling around its public API.