mksglu

Context Mode — Benchmark Results

Data community intermediate

SOURCE https://github.com/mksglu/context-mode

Description

> Benchmarked against **real outputs** from popular Claude Code MCP servers, Skills, and dev tools. > All fixtures captured from actual tool invocations — not synthetic data.

Installation

This entry records only its repository, not the path inside it, so there is no exact command to give. Open the source below and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

Repository README

This is the README for mksglu/context-mode, shared by 3 entries in this directory. It describes the repository, not this entry specifically.

Context Mode — Benchmark Results

Benchmarked against **real outputs** from popular Claude Code MCP servers, Skills, and dev tools. All fixtures captured from actual tool invocations — not synthetic data.

Overview

Metric	Value
Total scenarios	21
Tools benchmarked	`ctx_execute_file` (summarize) + `ctx_index`/`ctx_search` (knowledge retrieval)
Smart truncation	Head + tail preservation (60/40 split)
Total raw data processed	376 KB
Total context consumed	16.5 KB
Overall context savings	96%
Code examples preserved	100% (exact, not summarized)

Tool Decision Matrix

Data Type	Best Tool	Why
Documentation, API refs	`ctx_index` + `ctx_search`	Need exact code examples — not summaries
Skills prompts	`ctx_index` + `ctx_search`	Large prompts eat context; search on-demand
MCP tool signatures	`ctx_index` + `ctx_search`	Need exact tool names and parameters
Log files, test output	`ctx_execute_file`	Need aggregate stats, not raw lines
CSV data, analytics	`ctx_execute_file`	Need computed metrics
Build output	`ctx_execute_file`	Need error counts, not full logs
Browser snapshots	`ctx_execute_file`	Need page structure summary

Part 1: `ctx_execute_file` — Structured Data Processing

*Best for: logs, test output, CSV, build output — data where summaries are more useful than raw content.*

Scenario	Source	Raw Size	Context	Savings	Time
React useEffect docs	Context7	5.9 KB	261 B	96%	18ms
Next.js App Router docs	Context7	6.5 KB	249 B	96%	18ms
Tailwind CSS docs	Context7	4.0 KB	186 B	95%	18ms
Page snapshot (Hacker News)	Playwright	56.2 KB	299 B	99%	16ms
Network requests	Playwright	0.4 KB	349 B	13%	16ms
PR list (vercel/next.js)	GitHub	6.4 KB	719 B	89%	16ms
Issues (facebook/react)	GitHub	58.9 KB	1,139 B	98%	16ms
Test output (30 suites)	vitest	6.0 KB	337 B	95%	16ms
TypeScript errors (50)	tsc	4.9 KB	347 B	93%	16ms
Build output (100+ lines)	next build	6.4 KB	405 B	94%	16ms
MCP tools (40 tools)	MCP tools/list	17.0 KB	742 B	96%	15ms
Access log (500 requests)	nginx	45.1 KB	155 B	100%	17ms
Git log (150+ commits)	git	11.6 KB	107 B	99%	16ms
Analytics CSV (500 rows)	analytics	85.5 KB	222 B	100%	32ms

**Subtotal: 315 KB raw → 5.5 KB context (98% savings)**

Part 2: `ctx_index` + `ctx_search` — Knowledge Retrieval (FTS5 BM25)

*Best for: documentation, code examples, API references, Skills — content where you need EXACT text, not summaries.*

Scenario	Source	Raw Size	Search Result (3 queries)	Savings	Chunks	Code Blocks
Supabas