install-agent-skill
v1.0.1
Published
Enterprise-grade Agent Skill Manager with Antigravity Skills support, Progressive Disclosure detection, and semantic routing validation
Maintainers
Readme
Overview • Architecture • Commands
Installation
# Install agent skills
npx -y install-agent-skill dataguruin/agent-skills
# Install the complete governance pack
npx -y install-agent-skill dataguruin/agent-skills#coinpika-doctrine-pack
# Or install individual skills
npx -y install-agent-skill dataguruin/agent-skills#coinpika-pr-reviewer
npx -y install-agent-skill dataguruin/agent-skills#coinpika-commit-formatterOverview
Install Agent Skill is a production-grade CLI for orchestrating Agent Skills—modular, verifiable capability packs that extend AI agents without executing arbitrary code. It treats agent capabilities as managed dependencies, ensuring reproducibility and security across your engineering organization.
Key Value Proposition
- Zero-Execution Integrity: Skills are data, not scripts. No unauthorized code runs on your machine.
- Single Source of Truth: 'Smart Symlinks' keep your projects in sync with a central global registry.
- Semantic Routing: Automatically maps skills to the agents best suited to execute them.
Architecture
Progressive Disclosure
To prevent context window overflow, skills use a hierarchical loading system:
| Layer | Component | Description | Load Trigger |
|:------|:----------|:------------|:-------------|
| L1 | SKILL.md | Core directives & metadata | Always Active |
| L2 | doctrines/ | Specific behavioral policies | Context-Sensitive |
| L3 | resources/ | Deep documentation & specs | On-Demand |
Security & Verification
Your security is our priority.
- Merkle Hashing: Every installed skill is fingerprinted (
SHA-256) and stored in.skill-source.json. - Tamper Detection: Run
install-agent-skill doctorto verify that no files have been modified since installation.
Commands
| Command | Usage |
|:--------|:------|
| install <repo> | Interactive wizard to install skills from GitHub. |
| list | View installed skills, versions, and health status. |
| doctor | Diagnose configuration and integrity issues. |
| verify | Cryptographically verify installed assets against their manifest. |
| analyze <skill> | Inspect a skill's compatibility score and structure. |
| uninstall <skill> | Remove a skill and clean up symlinks/artifacts. |
License
MIT © DataGuruIn
