This repository is a curated collection of self-made, Agent Skills--compliant skills developed at the University of Tübingen as a joint effort of the Quantitative Biology Center (QBiC) and the Chair for Biomedical Data Science (BDS).
The primary purpose of this repository is to provide reusable, versioned skills that can be embedded into other Git projects (e.g., agent projects, automation repositories, internal tooling) while keeping validation and maintenance centralized.
- A collection of skills stored under
./skills/ - Skills intended to be fully agentskills.io compliant
- Continuous validation via GitHub CI using
sklintin strict mode - Optional agent governance via
AGENTS.md(see section below)
This repository is a skills library, not a full agent project.
.
├── skills/
│ ├── <skill-name>/
│ │ ├── SKILL.md
│ │ └── ...
├── .github/workflows/
├── AGENTS.md
└── README.md
This repository runs sklint in CI to ensure all skills conform to the Agent Skills specification. sklint is a Go CLI purpose-built for this — it is preferred over the reference implementation (skills-ref) bundled with the agentskills.io repository, which is intended for demonstration purposes only and is not suitable for production CI use.
Typical CI logic:
- Iterate over directories in
skills/* - Run
sklint --strict <skill-dir>
For installation instructions and full usage documentation, refer to the sklint repository.
git submodule add https://github.com/qbicsoftware/agent-skills vendor/tue-skills
git submodule update --init --recursive
Reference skills from:
vendor/tue-skills/skills/<skill-name>
Update later:
git submodule update --remote --merge
git subtree add --prefix vendor/tue-skills https://github.com/qbicsoftware/agent-skills main --squash
Update:
git subtree pull --prefix vendor/tue-skills https://github.com/qbicsoftware/agent-skills main --squash
git clone https://github.com/qbicsoftware/agent-skills
Copy or sync the skills/ directory as needed.
Depending on your tooling, you may:
- Point your agent tooling to the
skills/directory - Symlink selected skills into a local skills folder
Example:
mkdir -p .agents/skills
ln -s ../../vendor/tue-skills/skills/<skill-name> .agents/skills/<skill-name>
This repository contains an AGENTS.md to guide agents working inside
this repository.
When vendoring this repo (submodule/subtree), you may not want these instructions to apply to your project.
- Run agents from your project root (your own
AGENTS.mdtakes precedence). - Place the submodule in a directory like
vendor/orthird_party/and exclude it from agent scanning. - Override with your own root-level
AGENTS.md. - If supported by your tooling, explicitly disable agent instruction scanning for the vendored path.
- Add new skills under
skills/<skill-name>/ - Ensure Agent Skills compliance
- Run
sklint --strictlocally - CI must pass before merging
Maintained as a joint effort of:
- Quantitative Biology Center (QBiC), University of Tübingen
- Chair for Biomedical Data Science (BDS), University of Tübingen
Enforce clear, step-by-step, cause→effect reasoning and a single ordered plan. Use for tasks that benefit from structured analysis (requirements, debugging, implementation planning, checklists).
AGPL-3.0-only
- author:
Sven Fillinger - version:
1.0
This repository is lincensed under the AGPL-3.0 license