Metadata-Version: 2.5
Name: connexum-agent-standards
Version: 0.1.0b3
Summary: Standards coverage for any agent on any stack. Python twin of @connexum/agent-standards; one catalogue JSON, one conformance suite.
Author-email: "Connexum Network Inc." <support@my-cc.io>
License: SEE LICENSE IN LICENSE.md
Keywords: agent-security,ai-governance,hipaa,nist-ai-rmf,owasp
Requires-Python: >=3.11
Provides-Extra: testing
Requires-Dist: pytest>=8; extra == 'testing'
Description-Content-Type: text/markdown

# connexum-agent-standards

Python twin of `@connexum/agent-standards`: score any agent, on its first scan and every scan after, against OWASP LLM, OWASP Agentic and the My-CC runtime always, and against the frameworks whose My-CC pack is bound to it; frameworks that fit the agent's industry, data or regions but are not bound are documented as recommended, not scored. One catalogue JSON (`connexum_agent_standards/catalogue.json`, identical to the TypeScript package's) and one conformance suite drive both languages. A missing fact is "not checked", never a pass.

```python
from connexum_agent_standards import score_agent, applicability
report = score_agent({"industry": "healthcare", "dataClasses": ["phi"], "regions": ["us"], "tools": ["ehr.read_note"], "boundPacks": ["hipaa"]}, {"safe_harbor": True})
```
