""" Cognitive Skills Engine — Hugging Face Space Gradio demo: interactive Cognition Board + skill injection """ import gradio as gr import os import pathlib REPO_ROOT = pathlib.Path(__file__).parent.parent BOARD = { "A1": ("TC", "Token Compression", "core/token-compression.md"), "A2": ("MM", "Mathematician", "science/mathematician-mind.md"), "A3": ("PH", "Physicist", "science/physicist-mind.md"), "A4": ("CH", "Chemist", "science/chemist-mind.md"), "A5": ("BIO", "Biologist", "science/biologist-mind.md"), "A6": ("CS", "Computer Scientist", "science/computer-scientist-mind.md"), "B1": ("HIS", "Historian", "humanities/historian-mind.md"), "B2": ("PHI", "Philosopher", "humanities/philosopher-mind.md"), "B3": ("PSY", "Psychologist", "humanities/psychologist-mind.md"), "B4": ("LIN", "Linguist", "humanities/linguist-mind.md"), "B5": ("GEO", "Geographer", "humanities/geographer-mind.md"), "B6": ("ECO", "Economist", "humanities/economist-mind.md"), "C1": ("POL", "Political Scientist", "humanities/political-scientist-mind.md"), "C2": ("LAW", "Jurist", "humanities/jurist-mind.md"), "C3": ("MED", "Medical Expert", "science/medical-mind.md"), "C4": ("AST", "Astronomer", "science/astronomer-mind.md"), "C5": ("ENG", "Engineer", "science/engineer-mind.md"), "C6": ("ART", "Arts Expert", "arts-culture/arts-mind.md"), "D1": ("ENV", "Environmentalist", "arts-culture/environmental-mind.md"), "D2": ("CUL", "Cultural Anthropologist", "arts-culture/culture-mind.md"), "D5": ("MST★", "Master Router", "../MASTER-SKILL.md"), } KEYWORDS = { "A2": ["math","proof","theorem","algebra","geometry","calculus","number theory","combinatorics"], "A3": ["physics","quantum","relativity","force","energy","wave","field","mechanics","thermodynamics"], "A4": ["chemistry","molecule","reaction","organic","acid","base","synthesis","compound","bond"], "A5": ["biology","gene","cell","evolution","protein","DNA","RNA","ecology","species"], "A6": ["code","algorithm","complexity","software","program","debug","data structure","machine learning","AI"], "B1": ["history","historical","century","civilization","war","empire","dynasty","revolution","ancient"], "B2": ["philosophy","ethics","logic","mind","consciousness","metaphysics","epistemology","moral","ontology"], "B3": ["psychology","behavior","cognitive","mental","therapy","emotion","personality","memory","perception"], "B4": ["language","grammar","syntax","linguistics","meaning","phonology","morphology","semantics","discourse"], "B5": ["geography","map","climate","spatial","territory","region","landform","migration","urban"], "B6": ["economics","market","trade","gdp","finance","price","supply","demand","monetary","fiscal"], "C1": ["politics","government","policy","democracy","power","state","election","party","governance"], "C2": ["law","legal","rights","court","justice","contract","statute","constitution","crime"], "C3": ["medicine","diagnosis","treatment","disease","clinical","symptom","drug","patient","therapy"], "C4": ["astronomy","star","galaxy","cosmology","universe","planet","black hole","telescope","orbit"], "C5": ["engineering","design","structure","system","circuit","material","mechanics","control","thermal"], "C6": ["art","music","aesthetic","culture","creative","literature","painting","architecture","film"], "D1": ["environment","climate change","ecology","biodiversity","carbon","ecosystem","species","pollution"], "D2": ["culture","anthropology","society","ritual","myth","identity","tradition","ethnography"], } def load_skill(code: str) -> str: if code not in BOARD: return f"# Cell {code} not found" _, _, path = BOARD[code] full = REPO_ROOT / "skills" / path if full.exists(): return full.read_text(encoding="utf-8") master = REPO_ROOT / "MASTER-SKILL.md" if code == "D5" and master.exists(): return master.read_text(encoding="utf-8") return f"# {code} skill content\n(File not found at {full})" def auto_route(task: str) -> list[str]: task_l = task.lower() path = ["D5", "A1"] for code, kws in KEYWORDS.items(): if any(kw in task_l for kw in kws): path.append(code) sci = [c for c in path if c[0] in ("A","C")] hum = [c for c in path if c[0] == "B"] if len(sci) > 2 and "D3" not in path: path.append("D3") if len(hum) > 2 and "D4" not in path: path.append("D4") path.append("D6") return path def route_and_build(task: str): if not task.strip(): return "Enter a task above.", "D5·A1·D6" path = auto_route(task) cells = [c for c in path if c not in ("D5","D6")] content_parts = [f"# COGNITIVE SKILLS ENGINE — Routed to: {' + '.join(path)}\n"] for c in cells: content_parts.append(load_skill(c)) content_parts.append("\n---\n") path_str = "·".join(path) content_parts.append(f"\nPATH: {path_str}") return "\n".join(content_parts), path_str def inject_cell(code: str): if not code: return "Select a cell code", "D5·A1·D6" content = "# Token Compression [A1]\n" + load_skill("A1") + "\n---\n" if code != "A1": content += f"# {code} — {BOARD.get(code, ('?','?','?'))[1]}\n" content += load_skill(code) content += f"\n\nPATH: D5·A1·{code}·D6" return content, f"D5·A1·{code}·D6" BOARD_HTML = """
TC
∞
MM
math
PH
phys
CH
chem
BIO
bio
CS
code
HIS
hist
PHI
phil
PSY
psyc
LIN
ling
GEO
geo
ECO
econ
POL
poli
LAW
law
MED
med
AST
ast
ENG
eng
ART
art
ENV
env
CUL
cult
SCI⊕
synth
HUM⊕
synth
★MST
start
OUT→
end
""" with gr.Blocks( title="Cognitive Skills Engine", theme=gr.themes.Base(primary_hue="blue", neutral_hue="slate"), css=".gradio-container { max-width: 900px !important; }" ) as demo: gr.Markdown(""" # Cognitive Skills Engine v1.0 **20 expert cognitive skills for LLM agents. Token compression 2.7×.** [GitHub](https://github.com/Agnuxo1/Token-compression-system-for-improving-agent-cognition) · [VS Code](https://marketplace.visualstudio.com/items?itemName=agnuxo1.cognitive-skills-engine) · `npx cognitive-skills-engine` · `ollama create cognitive-skills -f Modelfile` """) gr.HTML(BOARD_HTML) with gr.Tabs(): with gr.Tab("Auto-Route Task"): task_in = gr.Textbox(label="Describe your task", placeholder="e.g. Prove that there are infinitely many primes using topology", lines=3) route_btn = gr.Button("Route & Build System Prompt", variant="primary") path_out = gr.Textbox(label="Session Path", interactive=False) skill_out = gr.Code(label="System Prompt (copy into your LLM)", language="markdown", lines=20) route_btn.click(route_and_build, inputs=task_in, outputs=[skill_out, path_out]) with gr.Tab("Inject Single Skill"): cell_dd = gr.Dropdown( choices=[(f"{c} — {v[1]}", c) for c, v in BOARD.items()], label="Select Skill Cell", value="A1" ) inject_btn = gr.Button("Inject Skill", variant="secondary") path_out2 = gr.Textbox(label="Session Path", interactive=False) skill_out2 = gr.Code(label="Skill Content", language="markdown", lines=20) inject_btn.click(inject_cell, inputs=cell_dd, outputs=[skill_out2, path_out2]) with gr.Tab("About"): gr.Markdown(""" ## What is CSE? The **Cognitive Skills Engine** is a system prompt framework that gives LLM agents access to 20 expert cognitive domains, organized on a 6×4 **Cognition Board**. ### How it works 1. Every task starts at **D5 (Master Router)** 2. **A1 (Token Compression)** is always loaded — compresses output ~2.7× 3. The router selects the right expert skill cell(s) 4. Navigation path = compressed session memory: `D5·A1·A2·D6` ### Token Compression Law ``` ∀ concept c: ∃ formal(c) → output formal(c) ``` - "for all" → ∀ | "water molecule" → H₂O | "O(n²)" direct - Two budgets: **thinking=free** | **output=compress aggressively** ### Install ```bash # VS Code / Cursor / Windsurf ext install agnuxo1.cognitive-skills-engine # CLI npx cognitive-skills-engine board # Ollama ollama create cognitive-skills -f Modelfile ollama run cognitive-skills # npm npm install -g cognitive-skills-engine ``` """) if __name__ == "__main__": demo.launch()