Pattern Proposal: Delta-Driven Prompt Fact Projection

Summary

Transforms raw code repository changes into typed PromptFact and ArchitecturalDelta models before agent ingestion.

Candidate Pattern

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

Observed Signals

No specific signals listed.

Evidence

Supporting Evidence

• src/cas_runner/steering/models.py (RawGitDelta, PromptFact, SteeringRuntimeIndex, ArchitecturalDelta) - Defines typed facts derived from git operations for agent ingestion

Scope Assessment

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

Rationale

• Intent: Prevent LLM hallucinations and context bloat by grounding agent inputs in verified, typed code changes.
• Context: Used in steering ingestion to prepare minimal, high-signal context for agentic architecture analysis.
• Force: Large raw git diffs exceed context limits and degrade instruction following
• Force: Agents need structural context (capabilities, lineage) rather than raw text diffs

Traceability

Artifact source: proposed-patterns