How yasirfarooq76/shakeela-begum-protocol-sbp- works
SBP is a lightweight Python prompting tool: a prompt is transformed through Negation, Interrogation, and Affirmation, and can also be run through model-backed workflows or benchmarked. The README documents the stages and benchmark modes; sampled source confirms a CLI, plug-in registry, provider implementations, and scoring/reporting helpers. Since excerpts omit much of the orchestration and transformation code, uncertain internal calls are not shown.
Main entry points: sbp.py Provider Base
Components 17
Protocol
- CLI
sbp.py - Prompt Transform
sbp.py - Negation
sbp.py - Interrogation
sbp.py - Affirmation
sbp.py - Model Run
sbp.py - Plug-in Registry
sbp.py
Model Providers
- Provider Base
sbp.py - Mock Provider
sbp.py - Anthropic Provider
sbp.py - OpenAI-compatible Provider
sbp.py
Evaluation
- Benchmark
sbp.py - Condition Judge
sbp.py - Score Aggregation
sbp.py - Report Output
sbp.py
Other components
- User
- Model Service
Connections 17
- User → CLI: submits prompt
- CLI → Prompt Transform: dispatches transform
- Prompt Transform → Negation: rewrites prompt
- Prompt Transform → Interrogation: rewrites prompt
- Prompt Transform → Affirmation: forms response
- CLI → Model Run: dispatches run
- Model Run → Provider Base: requests completion
- Anthropic Provider → Model Service: sends request
- OpenAI-compatible Provider → Model Service: sends request
- Mock Provider → Provider Base: implements
- Anthropic Provider → Provider Base: implements
- OpenAI-compatible Provider → Provider Base: implements
- CLI → Benchmark: dispatches bench
- Benchmark → Condition Judge: evaluates conditions
- Condition Judge → Provider Base: requests completion
- Condition Judge → Score Aggregation: returns scores
- Score Aggregation → Report Output: formats results