Skills

Catalog of UniverLab skills for AI agents.

The catalog is organized in three families, designed for dynamic, on-demand loading: skills are narrow and single-purpose so an agent (or a Canopy node pin) pulls only what the task needs. A Rust-loop skill in a Python session is context rot — install less, pin per role.

---

Behavior family

How the agent operates. execution-mindset is the base; architect-mindset and code-engineering stack on top per role, and tutor overrides its recommend-first rules for the length of a tutoring session. A skill applies only when loaded, so the agent's instruction file carries one line: load execution-mindset before any task.

execution-mindset

Default operating mode: the response sized to the request, indulgence named by its concrete forms (a spec that asks, a decision inside a spec, a bypass without a filed defect…), interview when two readings mean different work, search instead of recall for anything that changes yearly, the resolution model (model → first divergence → smallest decisive move → close the loop), recap to catch drift, verification before reporting.

architect-mindset

Behavior for design work: boundaries as contracts with violation signals, design for the process dying mid-step, reversible-vs-expensive decisions, and specs written for the cheapest executor.

code-engineering

Behavior for code work: structural judgment, pattern selection with restraint, maintainable multi-file changes, and debugging as the resolution model applied to code.

tutor

Maieutic tutoring for topics with conceptual depth: ceremony scaled to the request, interview by thematic axes, a Pareto roadmap the user restates, questions by default and neutral forks, minimal explanation only when the user is stuck, stages closed by recap.

git clone https://github.com/UniverLab/skills.git
# or as a submodule
git submodule add https://github.com/UniverLab/skills.git skills

---

Canopy tooling family

Only installed where the canopy binary exists — these are inert without the Canopy MCP server.

canopy-intelligence

The Project Intelligence Layer: pull workspace context at session start, persist facts/patterns/summaries with intelligence_upsert.

canopy-sync

Workspace coordination with the retired sync protocol's replacements: session_list, scheduled_prompts, and intelligence_upsert / intelligence_search — what to do instead of declaring, broadcasting, and reporting status.

canopy-capabilities

Orientation map for "what can I do here?": capabilities, starter automation ideas, and how to discover the live tool surface without trusting a static list.

canopy-graph-design

Design multi-step processes as reusable Canopy graphs: the tagged <spec> contract, design-expensive/execute-cheap model selection, ensembles and hooks, graph patterns, and the MCP tool playbook.

---

Tool skills

Bound to their CLI; installed automatically during Canopy setup only if the binary is available.

texforge

CLI to compile LaTeX documents to PDF without TeX Live, MiKTeX, or external dependencies. Uses tectonic as the internal engine. Requires texforge.

demostage

CLI for making terminal demos as reproducible code. Capture, record, export to gif/mp4. Requires demo.

# Manual installation of a single skill
npx skills add https://github.com/UniverLab/skills --skill texforge

# Or copy directly
cp -r skills/texforge ~/.agents/skills/

---

Pinning guide (Canopy graph nodes)

With Canopy's dynamic skills, pin per node role instead of installing globally:

Node role	Pin
Implementer	execution-mindset + code-engineering
Spec author / graph designer	architect-mindset + canopy-graph-design
Reviewer	execution-mindset (zero indulgence is the whole job)
Any Canopy-workspace session	canopy-intelligence, canopy-sync

---

Structure

Each skill follows this structure:

skill-name/
├── SKILL.md          # Main definition
├── README.md         # Full documentation
├── LICENSE           # MIT License
└── references/       # Reference documentation, loaded on demand
    ├── *.md          # Practical guides and examples

---

Usage

Reference these skills in your agent configuration:

skills:
  - name: execution-mindset
    path: skills/execution-mindset/SKILL.md
    always_active: true

  - name: code-engineering
    path: skills/code-engineering/SKILL.md
    triggers:
      - "write code"
      - "review code"
      - "design pattern"

  - name: architect-mindset
    path: skills/architect-mindset/SKILL.md
    triggers:
      - "design"
      - "architecture"
      - "write specs"

  - name: canopy-graph-design
    path: skills/canopy-graph-design/SKILL.md
    triggers:
      - "create graph"
      - "plan graph"
      - "orchestrate agents"

---

Contributing

Contributions welcome! Follow these guidelines:

1. One skill per directory
2. Clear documentation — Explain when and how to use the skill
3. Practical examples — Include code and real scenarios
4. MIT License — All contributions must be MIT
5. Progressive disclosure — Keep SKILL.md focused, move optional detail to references/
6. Narrow scope — a skill that only applies next to a binary or MCP server declares it with requires

---

Created with by JheisonMB and UniverLab