Skills¶

Cursor / agent instruction packs for BioModels tasks — ClawBio-style layout, lighter than full ClawBio skills (no separate Python runner required; uses PraisonAI toolsets).

Catalog: skills/catalog.json

Available skills (16)¶

Skill	Use when
Search BioModels	Find models by keyword
Get model info	Metadata for one ID
Load BioModels SBML	Download SBML file
Simulate model	Time-course or steady state
Ask question	Plain-English Q&A about a model
Steady state	Equilibrium concentrations
Parameter scan	Sweep a parameter range
SBML summarise	Structure overview
SBML validate	Check SBML file
Compare models	Two models side by side
Rank models	Sort search results
Trust scorecard	Curation quality
Query article	Linked publications
SED-ML parse	Read experiment definitions
SysBio orchestrator	Multi-step pipeline
Repro enforcer	Export study bundle

Developer: how skills are registered

Catalog: skills/catalog.json · Template: skills/SKILL-TEMPLATE.md

See sections below for adding new skills.

Legacy table (core five)¶

Skill	ID	Use when
Search BioModels	`search-models`	Find curated models for a pathway or keyword
Load BioModels SBML	`load-biomodel`	Download and inspect SBML on disk
Simulate Model	`simulate-model`	Run BASICO time-course or steady state
SysBio Orchestrator	`sysbio-orchestrator`	Multi-step discovery → simulate pipeline
Reproducibility Enforcer	`repro-enforcer`	Export repro bundles with checksums

PraisonAIBio vs ClawBio skills¶

	PraisonAIBio	ClawBio (`~/ClawBio`)
Purpose	Systems biology / BioModels / SBML	Genomics, WES, RNA-seq, etc.
Implementation	Markdown + PraisonAI `@tool` functions	Often `SKILL.md` + Python script + tests
Registration	`skills/catalog.json` via `generate_catalog.py`	`catalog.json` + `clawbio.py` aliases
Scaffold	Copy `SKILL-TEMPLATE.md`	Use `skill-builder` meta-skill
Bridge	`mcp/clawbio-bridge/` (v2)	N/A

For heavy genomics pipelines, use ClawBio directly. For BioModels discovery and simulation, use PraisonAIBio skills + toolsets.

How to add a new skill¶

1. Create the folder¶

Use a kebab-case id (matches directory name):

mkdir -p skills/my-skill-id
cp skills/SKILL-TEMPLATE.md skills/my-skill-id/SKILL.md

2. Edit `SKILL.md`¶

Required sections (see template):

Title — first line # Title (used as display name in catalog)
When to use — trigger scenarios for the agent
Tools — list tools with parameters (agents only see parameters, not env vars)
Toolset — which prebuilt toolset to attach (biomodels-readonly, simulation, …)
Example — minimal Agent snippet loading this skill as instructions

Optional: prerequisites ([simulation] extra), demo model ID.

Tip

Copy an existing skill as a starting point: cp -r skills/search-models skills/my-skill-id

3. Regenerate the catalog¶

python scripts/generate_catalog.py

This scans skills/*/SKILL.md and rewrites skills/catalog.json.

4. Validate¶

python scripts/validate_repo.py   # skills section must stay OK
python -c "import praisonai_bio"
praisonai-bio validate check

5. Test with an agent¶

import praisonai_bio
from praisonaiagents import Agent

agent = Agent(
    name="test",
    instructions=open("skills/my-skill-id/SKILL.md").read(),
    toolsets=["biomodels-readonly"],  # match your SKILL.md
    llm="gpt-4o-mini",
)
print(agent.start("Your test prompt"))

6. Document and submit¶

Add a row to the table above (this page) if the skill is public
Open a pull request with SKILL.md, updated catalog.json, and any tests

Optional extras (like ClawBio, but not required for PraisonAIBio):

skills/my-skill-id/
├── SKILL.md           # required
├── examples/          # optional sample prompts or output
└── tests/             # optional pytest for helper scripts

Use with Cursor¶

Copy or symlink the skill folder into your Cursor skills directory, or
Reference the repo path in chat:

Read skills/search-models/SKILL.md and search BioModels for MAPK models.

Use with PraisonAI agents¶

Skills supply instructions; toolsets supply capabilities:

import praisonai_bio
from praisonaiagents import Agent

agent = Agent(
    name="scout",
    instructions=open("skills/search-models/SKILL.md").read(),
    toolsets=["biomodels-readonly"],
    llm="gpt-4o-mini",
)
agent.start("Find curated p53 models")

Always run import praisonai_bio first.

YAML workflows¶

Point agent instructions at a skill file in .praisonai/agents/ or inline in workflow YAML:

agents:
  scout:
    role: BioModels Scout
    instructions_file: skills/search-models/SKILL.md
    toolsets:
      - biomodels-readonly

(If your PraisonAI version uses instructions: instead of instructions_file:, paste or load the markdown content.)