Security-reviewed, reusable assets for AI agents used by the team.
| Directory | Purpose |
|---|---|
skills/ |
Executable agent skills in the open Agent Skills layout |
rules/ |
Reusable instruction and policy text that a project can adopt |
memory/ |
Sanitized, durable knowledge that is safe to share publicly |
This repository is public. Never add credentials, customer data, personal data, private conversation exports, or internal-only project history.
List the available skills:
npx skills add wunderforge/shared-agent-assets --listInstall one skill for a supported agent:
npx skills add wunderforge/shared-agent-assets --skill <skill-name> --agent codexThe same repository can target Claude Code, Cursor, and other clients supported
by the skills CLI.
- Create a branch and add one coherent asset or change.
- Run
./scripts/check.ps1on Windows or./scripts/check.shon macOS/Linux. - Open a pull request using the repository template.
- Merge only after owner review and the required CI gate pass.
Every skill is scanned with
NVIDIA SkillSpector. The gate is
fail-closed: only a SAFE recommendation is accepted.
See CONTRIBUTING.md for the asset contracts and review flow.