015-context-engineering

015-context-engineering is a cursor rule for coding agents from d-padmanabhan/agent-engineering-handbook. It costs 815 tokens per session, scanned A, original, MIT.

A set of rules for giving coding agents the right information at the right time, while keeping that information concise and reusable.

In plain words
What is it for?
It helps agents pack goals, constraints, plans, and current status into prompts, retrieve files or logs when needed, and save short summaries for later.
Why use it?
It reduces confusion caused by overloaded prompts and helps work continue across sessions without rereading the whole project.

Cursor rule

Install

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.

agentmods
npx agentmods add rules/d-padmanabhan/agent-engineering-handbook/015-context-engineering
Clone the repo
git clone --depth 1 https://github.com/d-padmanabhan/agent-engineering-handbook

Wrote 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.

agentmods badge for 015-context-engineering

README.md
[![agentmods](https://agentmods.dev/badge/rules/d-padmanabhan/agent-engineering-handbook/015-context-engineering.svg)](https://agentmods.dev/rules/d-padmanabhan/agent-engineering-handbook/015-context-engineering)
Your own site
<a href="https://agentmods.dev/rules/d-padmanabhan/agent-engineering-handbook/015-context-engineering"><img src="https://agentmods.dev/badge/rules/d-padmanabhan/agent-engineering-handbook/015-context-engineering.svg" alt="Measured on agentmods" height="20"></a>
Per session 815 This file is loaded in full into every session.
When invoked 815 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00815 $0.00815
Opus 5 $0.00407 $0.00407
Sonnet 5 $0.00163 $0.00163
Haiku 4.5 $0.00081 $0.00081

Measured 3d ago against content hash a91e6d1f4384, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

015-context-engineering 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 3d 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.

rules/015-context-engineering.mdc · 116 lines

How it starts

The opening of the file, as written. The whole thing — 116 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Context Engineering for Coding Agents

This rule strengthens agent performance by making context intentional, minimal, and reproducible.

[!NOTE] This rule complements 010-workflow.mdc (phases + context files) and 100-core.mdc (anti-slop guardrails).


Goals

  • Keep the agent’s working context small and high-signal
  • Make work resumable across sessions without “reloading the world”
  • Prefer retrieval over memory: fetch exactly what you need, when you need it

Mental Model: Three Layers of Context

  • Working context (volatile): what the model sees right now
  • Repo context (retrievable): files, diffs, logs, docs that can be re-read on demand
  • Durable context (portable): short summaries + decisions captured in tmp/ context files

[!IMPORTANT] If it’s important and will matter later, it must be captured in durable context.


Prompt Packing (What to Carry vs. What to Fetch)

Always carry (paste/recap)

  • Goal: one sentence
  • Non-goals: 1-3 bullets
  • Constraints: security, tooling, performance, “don’t refactor”, etc.
  • Current plan: 3-7 steps max
  • Current state: what’s done / what’s blocked
  • Critical artifacts: exact file paths, key identifiers, and error messages (verbatim)

Never carry (fetch instead)

  • Entire files or long logs “just in case”
  • Repeated restatements of requirements
  • Unbounded command output without a reason

Retrieval Strategy (Search → Read Narrow → Cite)

Use a tight loop:

  1. Search for the smallest set of candidate files (prefer repo-local sources first)
  2. Read only the minimal sections needed to decide or implement
  3. Cite exact file paths / relevant snippets in your reasoning and updates

[!TIP] Prefer “retrieve just-in-time” over “load everything up-front”. It is faster and more accurate.


Compaction Strategy (Frequent Intentional Compaction)

After any meaningful milestone (plan approved, a module completed, a bug fixed), compact:

Read the full file on GitHub · 116 lines

Changes

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.

  1. 3d ago First seen · 116 lines · 815 tokens per session scan A a91e6d1f4384

Subscribe to this mod's changes

015-context-engineering is a cursor rule published in the GitHub repository d-padmanabhan/agent-engineering-handbook (16 stars, last pushed 4d ago), licensed MIT. It adds 815 tokens to every session, about $0.0041 per session on Opus 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 2026-08-30.