Analysis toolkit for agentic context management research. Triage papers and tools into structured reference summaries and REVIEWED.md entries following the agentic-context repo conventions.
93
93%
Does it follow best practices?
Impact
Pending
No eval scenarios have been run
Advisory
Suggest reviewing before use
Used when promoting a tool from REVIEWED.md to a standalone deep dive (ANALYSIS-<slug>.md).
title: "Analysis — <Tool Name>"
date: YYYY-MM-DD
type: analysis
tool:
name: "<Tool name>"
repo: "<GitHub URL>"
version: "<vX.Y.Z / commit SHA>"
language: "<Python / TypeScript / Rust / etc>"
license: "<license>"
source:
- "references/<tool-reference-summary>.md"
- "tools/<repo-name>/"1.1 Problem it solves
1.2 Architecture overview
1.3 Context representation (storage primitives)
1.4 Injection mechanism (what gets injected, priority logic, system prompt structure)
1.5 Compression / summarisation (trigger → method → fidelity → reversibility)
1.6 Eviction / overflow handling (detection → policy → archival)
1.7 Session lifecycle (init / per-turn / checkpoint / cross-session continuity)
1.8 What the tool explicitly does not cover
2.1 Self-reported metrics (as reported)
2.2 Independent verification (code audit / benchmark cross-check)
2.3 Strengths
2.4 Limitations / risks / missing details
2.5 Operational complexity (deployment dependencies, failure modes, observability)
3.1 Comparison to adjacent tools
3.2 Mapping to repo themes (compression / tiered loading / token budgeting / injection)
3.3 Concrete takeaways (actionable)
Always record scope fit in the analysis: in scope / borderline / out of scope — with reasoning.