Metadata-Version: 2.5
Name: spangloss
Version: 0.0.1
Summary: Security labels for AI agent telemetry: typed, probabilistic annotations on OpenTelemetry spans.
Project-URL: Homepage, https://github.com/spangloss/spangloss
Author: Phil DiCorpo
License-Expression: Apache-2.0
License-File: LICENSE
Keywords: ai-agents,mcp,observability,opentelemetry,security
Classifier: Development Status :: 1 - Planning
Classifier: Programming Language :: Python :: 3
Classifier: Topic :: Security
Classifier: Topic :: System :: Monitoring
Requires-Python: >=3.12
Description-Content-Type: text/markdown

# spangloss

Dr. Pangloss assumed everything was for the best. Spangloss checks.

Spangloss adds security labels to AI agent telemetry. It asks a decision model narrow, typed questions about each OpenTelemetry span — what the call does, what data it touches, whether a tool result contains instructions aimed at the agent — and writes the answers back as `sec.*` attributes with probabilities. Detections and audits can then run over labels instead of raw prompts.

**Status:** early development. The first public release, with a published label taxonomy and per-label evaluation results, is planned for late November 2026.
