Skip to content

Awesome list for AI agent harness engineering: tools, patterns, evals, memory, MCP, permissions, observability, and orchestration.

Add Governance Decay (arXiv 2606.22528) to Context Delivery & Compaction

ai-boostf106c46ca2f3
diff --git a/README.md b/README.md
index 00a744e..087de02 100644
--- a/README.md
+++ b/README.md
@@ -280,6 +280,7 @@ Harness components organized by the problem they solve, not by vendor.
- [MemArchitect: A Policy-Driven Memory Governance Layer](https://arxiv.org/abs/2603.18330) — Proposes a governance layer that decouples memory lifecycle management (decay, conflict resolution, privacy enforcement) from model weights, directly addressing the "zombie memory" problem: outdated facts sitting in the context window that only a harness-level eviction policy — not the model — can remove.
- [Codified Context: Infrastructure for AI Agents in a Complex Codebase](https://arxiv.org/abs/2602.20478) — Production-validated architecture (283 sessions, 108k-line codebase) built on three components: a "hot-memory constitution" encoding conventions and multi-agent coordination protocols, 19 domain-specialist agents, and a "cold-memory knowledge base" of 34 on-demand specification documents. The empirical data distinguishes what must live in always-on context from what should be retrieved on demand — the most concrete published guidance for scaling cross-session memory in a large codebase.
- [Facts as First Class Objects: Knowledge Objects for Persistent LLM Memory](https://arxiv.org/abs/2603.17781) — Identifies three production failure modes of in-context memory at scale: capacity overflow at ~8,000 facts, 60% fact destruction during compaction, and 54% behavioral drift from constraint erosion across cascaded summarizations. Proposes Knowledge Objects (hash-addressed discrete fact tuples) achieving 100% accuracy at 252× lower cost than in-context storage — the quantitative case for moving persistent facts out of the context window into a structured retrieval layer rather than managing them through prompt engineering.
+- [Governance Decay: How Context Compaction Silently Erases Safety Constraints in Long-Horizon LLM Agents](https://arxiv.org/abs/2606.22528) — June 2026 paper establishing context compaction as a first-class governance surface: the ConstraintRot benchmark (1,323 episodes, seven model families) shows constraint-violation rates rising from 0% with the policy in full context to 30% after compaction (59% for some models), and a Compaction-Eviction Attack that biases the summarizer to drop legitimate policies defeats every evaluated model. The training-free Constraint Pinning mitigation — quarantining governance constraints from lossy compaction — restores violations to 0%, making it the concrete design rule that safety-critical instructions must live outside the compactable context region.
- [Recoverability Has a Law: The ERR Measure for Tool-Augmented Agents](https://arxiv.org/abs/2601.22352) — Formal framework for measuring how well agents recover from tool failures. Defines Expected Recovery Regret (ERR) as a metric for harness design: the cost of recovering from stochastic failures in downstream tasks. Critical for assessing reliability of production harnesses where tool calls occasionally fail but agents must continue functioning.
- [MAGMA: Multi-Graph Agentic Memory Architecture](https://arxiv.org/abs/2601.03236) — Represents agent memory across four orthogonal semantic, temporal, causal, and entity graphs, enabling policy-guided retrieval over relational views. Outperforms MemGPT on long-horizon reasoning benchmarks by 18.5% accuracy improvement. The multi-graph abstraction lets harness engineers compose different retrieval strategies for different task phases — a concrete architecture for memory that scales beyond single-view approaches.
- [GAAMA: Graph Augmented Associative Memory for Agents](https://arxiv.org/abs/2603.27910) — Hybrid memory system blending graph traversal with semantic similarity through additive scoring; graph augmentation improves retrieval over embedding-only approaches for long-horizon reasoning. Practical alternative to full multi-graph systems when adding structure to existing vector-based memory is sufficient.