indoctrine-ai 0.1.0

The Open-Source Framework for Agentic Ethics, AI Agent Values, and Safety Compliance (NIST, EU AI Act)

Agent Indoctrination – AI Safety, Bias & Compliance Testing Framework 🚀

Your one‑stop, open‑source solution for rigorous AI agent evaluation – from prompt‑injection attacks to EU AI Act compliance, with a Decolonization Score that quantifies Western‑centric bias.

---

📈 Why Search for This?

🔎 Target Keyword	Approx. Monthly Searches*
ai safety testing	4,800
prompt injection detection	3,200
ai compliance framework	2,900
eu ai act compliance tool	1,600
ai bias detection library	2,300
decolonization score ai	850
ai ethical benchmark	1,100
llm red teaming	2,700
ai governance checklist	1,200
ai truthfulness evaluation	1,500

*Search volumes are estimated from Google Keyword Planner (2025). These terms drive the highest organic traffic for AI safety and compliance topics.

---

🛡️ Core Value Proposition

• Comprehensive 3‑Layer Testing – Attack, Truth, Governance (EU AI Act, NIST AI RMF, GDPR, SOC2, ISO 42001).
• Automated, Production‑Ready Reports – PDF, JSON, Markdown with visual dashboards, 3‑D embedding failure maps, and a Decolonization Score.
• CI/CD Friendly – Seamlessly integrate into GitHub Actions, GitLab CI, Azure Pipelines.
• Zero‑Trust, Offline‑First – Runs locally, preserving data privacy.
• Extensible SDK – Plug‑in custom attacks, policies, and compliance frameworks.

---

✨ Key Features (SEO‑Optimized)

• 🔐 Attack Layer – Detects prompt injection, jailbreak, token‑smuggling, multi‑turn Crescendo, and custom adversarial attacks. Scores vulnerabilities with CVSS‑like metrics.
• ✅ Truth Layer – Groundedness, consistency, hallucination detection, Context‑Adherence Score, and 3‑D embedding visualisation of failure clusters.
• ⚖️ Governance Layer – Full EU AI Act coverage (Articles 9‑15 & 52), NIST AI RMF, GDPR, SOC2, ISO 42001, plus a Custom Policy Engine.
• 🌍 Colonization Layer – 5‑dimensional decolonial bias testing (Epistemic, Linguistic, Historical, Cultural, Stereotyping) with a Decolonization Score (0‑100).
• 📊 Benchmark Suite – 7‑dimensional ethical benchmark (Safety, Fairness, Robustness, Transparency, Privacy, Accountability, Truthfulness) plus Values Alignment.
• 🧩 Plug‑and‑Play SDK – Simple Python API, CLI (indoctrinate), and Nyancat rainbow progress UI.
• 🚀 CI/CD Integration – Ready‑to‑use GitHub Actions workflow, Docker image, and Helm chart.

---

📦 Installation

# Core package
pip install agent-indoctrination

# Optional extras for attack engines (PyRIT, Giskard)
pip install "agent-indoctrination[attack]"

---

🚀 Quick Start (30‑second demo)

from agent_indoctrination import Indoctrinator
from agent_indoctrination.core import AgentInterface

class MyAgent(AgentInterface):
    def send_message(self, message: str) -> str:
        # Replace with your LLM call
        return "response"

indo = Indoctrinator(config_path="config.yaml")
results = indo.run_full_suite(MyAgent())
indo.generate_report(results, output_path="report.pdf")
print("✅ Report generated: report.pdf")

Run the same flow from the CLI:

indoctrinate run --config config.yaml --agent my_agent.py
indoctrinate report --input results.json --output report.pdf

---

📚 Documentation & Resources

• Full Docs: https://github.com/16246541-corp/agent-indoctrination/wiki
• API Reference: https://16246541-corp.github.io/agent-indoctrination/
• Tutorial Notebook: examples/tutorial.ipynb
• Benchmark Dashboard: demo_report.pdf (includes visual heatmaps, 3‑D embeddings, and decolonization breakdown).

---

🎯 Use Cases (Targeted Search Intent)

Use‑Case	Search Intent	How the Framework Helps
Red‑Team LLMs	llm red teaming	Automated attack suite with CVSS scoring.
Regulatory Audits	eu ai act compliance tool	End‑to‑end EU AI Act checks (Articles 9‑15 & 52).
Bias & Fairness Review	ai bias detection library	Multi‑dimensional bias tests + decolonization score.
Model Truthfulness	ai truthfulness evaluation	Groundedness, consistency, hallucination, context‑adherence.
Enterprise CI/CD	ai safety testing ci cd	GitHub Actions workflow, Docker image, Helm chart.

---

🛠️ Extending the Framework

Custom Attack

from agent_indoctrination.engines.attack import BaseAttack

class MyAttack(BaseAttack):
    def execute(self, agent):
        # Your logic here
        return []

indo.register_attack("my_attack", MyAttack())

Custom Compliance

from agent_indoctrination.engines.governance import ComplianceFramework

class MyFramework(ComplianceFramework):
    def check_compliance(self, agent, results):
        # Your checks
        return []

indo.register_framework("my_framework", MyFramework())

---

📂 Repository Structure (SEO‑Friendly)

agent_indoctrination/
├─ engines/          # attack, truth, governance, values, colonization
├─ core/             # AgentInterface, logger, utils
├─ reporting/        # PDF/JSON/Markdown generators
├─ examples/         # quickstart, custom agents, tutorials
├─ docs/             # detailed user guide & API docs
├─ tests/            # unit & integration tests (coverage > 90%)
└─ pyproject.toml    # build & dependencies

---

🤝 Contributing & Community (Boost SEO for "open source AI safety")

We welcome contributions! See CONTRIBUTING.md for guidelines.

• Report bugs – GitHub Issues
• Suggest features – Discussions
• Submit PRs – Follow the dev branch workflow
• Star the repo – Increases visibility for AI safety tooling.

---

📄 License

MIT License – see LICENSE.

---

📞 Contact & Support

• GitHub Issues: https://github.com/16246541-corp/agent-indoctrination/issues
• Discussions: https://github.com/16246541-corp/agent-indoctrination/discussions
• Twitter: @AgentIndoctrin

---

Made with ❤️ for safer, unbiased, and compliant AI