Shared Cursor agent skills and shared Cursor rules for DataKnifeAI repositories.
skills/
<skill-name>/
SKILL.md
scripts/ # optional
rules/
dataknifeai-common.mdc
templates/ # project override examples
- Create
skills/<skill-name>/SKILL.mdwith YAML frontmatter (name,description) and clear instructions. - Open a PR; keep scope to one new skill or one focused change per PR when possible.
Org-wide agent defaults live in rules/dataknifeai-common.mdc. Vendor a copy into each product repo at .cursor/rules/dataknifeai-common.mdc and commit it; add thin project overrides beside it. Normal clones then pick up the rules with no install step — see rules/README.md.
When the org baseline changes, refresh by re-copying the canonical file from this repo into each product repo’s .cursor/rules/dataknifeai-common.mdc.
Cursor loads project skills from .cursor/skills/<skill-name>/ in the workspace (or user skills from ~/.cursor/skills/).
From a checkout of this repository:
make help # list targets (default)
make install # ./.cursor/skills (symlinks)
make install-user
make dry-run
make list # skill directory names
make validate # skills + rules
make clone-install # optional: CLONE_URL=… CACHE=…
make clean-project # rm -rf ./.cursor/skillsSubset: make install ONLY=gh-cli,kubectl-cli.
From a checkout of this repository:
./scripts/install-skills.sh # ./.cursor/skills (symlinks)
./scripts/install-skills.sh --user # ~/.cursor/skills
./scripts/install-skills.sh --only gh-cli,kubectl-cli
./scripts/install-skills.sh --copy # copy trees instead of symlinksBootstrap skills from GitHub without cloning the full repo yourself (shallow clone to the cache, then symlink into ./.cursor/skills):
bash -c 'curl -fsSL https://raw.githubusercontent.com/DataKnifeAI/agent-skills/main/scripts/install-skills.sh | bash -s -- --clone https://github.com/DataKnifeAI/agent-skills.git'Use --clone with an internal mirror URL if your deployment mirrors this repo. Override the cache directory with --cache DIR. Run ./scripts/install-skills.sh --help for all options.
If you prefer not to use the script: clone or submodule this repo, then symlink each skills/<name>/ directory into .cursor/skills/<name>/, or install under ~/.cursor/skills/. For rules, copy rules/dataknifeai-common.mdc into the product repo’s .cursor/rules/ and commit it.
Ensure glab, gh, coder, kubectl, and/or rancher are on PATH in the workspace image when those skills apply; the skills document CLI usage only.
This repository is licensed under the Apache License 2.0 — see LICENSE.
