Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add roedyrustam/vibes-plug --skill context-window-engineergit clone --depth 1 https://github.com/roedyrustam/vibes-plugWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/roedyrustam/vibes-plug/context-window-engineer)<a href="https://agentmods.dev/skills/roedyrustam/vibes-plug/context-window-engineer"><img src="https://agentmods.dev/badge/skills/roedyrustam/vibes-plug/context-window-engineer/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/roedyrustam/vibes-plug/context-window-engineer"><img src="https://agentmods.dev/badge/skills/roedyrustam/vibes-plug/context-window-engineer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00124 | $0.03747 |
| Opus 5.5 | $0.00050 | $0.01499 |
| Sonnet 5.5 | $0.00025 | $0.00749 |
| Haiku 4.5 | $0.00012 | $0.00375 |
Grade A, and why
context-window-engineer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 303 lines — stays where its author put it; the contents beside it link to each section on GitHub.
context-window-engineer — vibes-plug Skill
English
Orchestration & Integration
Connects and orchestrates with kv-cache-prefix-optimizer, llm-finops-router, frontier-ai-models-expert, gemini-agent-booster, ai-llm-integration-expert, vercel-ai-sdk-expert, adaptive-model-cascade, session-memory-manager, and deep-research-analyst to ensure context windows are utilized efficiently across all AI-powered workflows.
Description
Production guide for engineering, partitioning, and optimizing ultra-large context windows (128k → 2M+ tokens) across frontier models. Prevents the most common failure mode in 2026 AI applications: dumping raw data into context without structure, causing degraded recall, wasted tokens, and hallucination amplification in the "lost in the middle" zone.
Trigger Conditions
Activate this skill when:
- Building AI agents or applications that consume >50k tokens of context per request.
- Experiencing degraded model recall, "lost in the middle" failures, or hallucinations in long-context scenarios.
- Architecting Retrieval-Augmented Generation (RAG) systems that need to decide between retrieval vs. long-context stuffing.
- Optimizing token costs for applications with large system prompts, tool registries, or codebase context maps.
- Designing multi-turn agentic loops where context accumulates across dozens of turns.
Core Concepts & Patterns
1. Context Window Capacity Map (October 2026)
| Model | Max Context | Effective Recall Zone | Cost per 1M Input Tokens | Cache Discount |
|---|---|---|---|---|
| Gemini 4 Pro | 2,097,152 | ~1.8M (reliable) | $1.25 | 75% (explicit cache) |
| Gemini 4 Flash | 1,048,576 | ~900k | $0.075 | 75% |
| Claude 5.5 Opus | 500,000 | ~450k | $15.00 | 90% (prompt cache) |
| Claude 5.5 Sonnet | 200,000 | ~180k | $3.00 | 90% |
| GPT Astra 6 | 256,000 | ~220k | $5.00 | 50% (prefix cache) |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 8d ago Changed 068750eaef93
- 13d ago First seen · 303 lines · 124 tokens per session scan A 72c6a284d4f2
context-window-engineer is a skill published in the GitHub repository roedyrustam/vibes-plug (75 stars, last pushed yesterday), licensed MIT. It adds 124 tokens to every session and 3,747 once invoked, about $0.0005 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-10-04.
Other skills, from other repositories
rag-and-memory
Patterns for Retrieval-Augmented Generation (RAG) and agent memory systems. Retrieves only relevant context, prevents context bloat, and maintains coherent state across sessions.
ai-product
Every product will be AI-powered. The question is whether you'll build it right or ship a demo that falls apart in production.
llamaindex
Expert LlamaIndex assistance covering RAG (Retrieval-Augmented Generation), vector store indexes, document chunking, and query engines. Use when connecting LLMs to private enterprise data and document stores.
neo4j
Expert Neo4j graph database assistance covering Cypher query language, node/edge labeling, graph algorithms, and APOC. Use when modeling connected data, recommendation engines, fraud detection, or knowledge graphs.
ai-pipeline-orchestration
Orchestrate AI/ML pipelines for data ingestion, model training, batch inference, and RAG indexing using Prefect, Airflow, or Dagster.
ai-ml
AI and machine learning workflow covering LLM application development, RAG implementation, agent architecture, ML pipelines, and AI-powered features.