Pattern Proposal: Prompt Fact Projection Indexing

Summary

Transforms raw Git changes and capability records into discrete, verified PromptFacts organized in an index to ground LLM context.

Candidate Pattern

• ID: prompt-fact-projection
• Classification: intelligence
• Maturity: established
• Type: implemented

Observed Signals

No specific signals listed.

Evidence

Supporting Evidence

• src/cas_runner/steering/models.py (PromptFact, SteeringRuntimeIndex, CapabilityRecord, ArchitecturalDelta) - Models data contract for compiling git changes into structured facts for prompts

Scope Assessment

• Evidence Strength: high
• Confidence: high
• Applies to: ideas-to-life-cas

Rationale

• Intent: Ground agent operations in strictly verified architectural constraints and delta facts without context bloat.
• Context: Used in cas_runner's steering module when preparing contextual input for architecture validation and agentic updates.
• Force: Unbounded context windows lead to loss of instruction adherence
• Force: Agents need up-to-date knowledge of evolving capabilities without manual prompt crafting

Traceability

Artifact source: proposed-patterns