- Home
- Skills
- Agents & Orchestration
- context engineering
Works with the AI Agents you already use
context engineering
Architect, audit, and optimize context systems for coding agents, MCP workflows, and RAG pipelines.
$12
context engineering
Example session with this skill installed
My Claude Code agent keeps ignoring our Vitest patterns and trying to use Jest. Here is my current CLAUDE.md and a sample test file. Audit this and fix the context drift.
- Read your context and instructions
- Compiled the context engineering
I have audited your CLAUDE.md and the provided test sample. The primary issue is context confusion: your instructions mention "standard JavaScript testing" which the model is defaulting to Jest, while your file structure uses .test.ts with Vitest-specific globals.
Context Audit & Repair for Vitest Migration
| Finding | Impact | Recommended Fix |
|---|---|---|
| Ambiguous Tooling | Agent defaults to Jest globals (describe, it) | Explicitly define Vitest as the sole test runner in CLAUDE.md. |
| Missing Commands | Agent guesses npm test which fails | Add specific Vitest CLI commands to the Test Commands section. |
| Pattern Starvation | No example of a Vitest mock in context | Add a concise "Golden Example" block for component mocking. |
| Context Poisoning | Stale reference to setupTests.js | Remove legacy Jest setup references from the project rules. |
Updated CLAUDE.md Section
## Testing Conventions
- Use Vitest for all unit and integration tests.
- Test files must reside in `src/**/__tests__/*.test.ts`.
- Use `vi.mock()` for dependency injection; do not use `jest.mock()`.
- Run specific test: `npx vitest run <path>`
- Run all tests: `npm run test:ci`
## Golden Example (Vitest)
import { vi, test, expect } from 'vitest';
import { service } from './service';
test('logic', () => {
const spy = vi.spyOn(service, 'call');
// ... test logic
});
Next steps
- Replace the
## Testingsection in your rootCLAUDE.mdwith the updated block above. - Run
npx vitest runand provide the output to the agent to verify it recognizes the environment. - If the agent still confuses syntax, we should implement a
.cursorrulesfile for editor-level enforcement.
Connects securely to your tools. The creator never sees your data.
About this skill
The problem
Coding agents and RAG pipelines often suffer from context collapse, hallucinated APIs, or token waste. Developers struggle to manage the boundaries between static instructions, dynamic project rules, and retrieved knowledge, leading to degraded agent performance over long sessions.
What it does
- Designs structured
CLAUDE.md,AGENTS.md, and.cursorrulesfiles to enforce project conventions. - Audits existing context setups to identify starvation, poisoning, or conflicting instructions.
- Creates multi-agent handoff packets that preserve state, authority, and evidence across transitions.
- Develops retrieval strategies and memory architectures for RAG systems to minimize noise.
- Builds context compression and eviction plans to stay within token budgets without losing critical detail.
Frameworks & tools
Claude Code, Cursor, GitHub Copilot, MCP (Model Context Protocol), RAG architectures, and agentic workflows.
Why this beats prompting it yourself
Writing a good prompt is not enough for complex agents. This skill applies information architecture to context, separating static rules from dynamic state. It prevents context drift by establishing formal quality gates and handoff contracts that work across multi-turn loops.
Use cases
- Setting up a new repository for Claude Code or Cursor with strict architectural rules.
- Diagnosing why an agent is missing local testing conventions or hallucinating legacy code.
- Defining the data contract and escalation paths for a multi-agent orchestration system.
- Optimizing a RAG pipeline's context window by ranking and compressing retrieved artifacts.
Known limitations
Does not modify live source code directly without a prior design review. Requires manual input of environment-specific constraints like token limits or write permissions.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
Requires explicit environment details (e.g., Claude Code vs Cursor). Does not perform code edits directly. User must provide repo structure or existing rules for audits.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 2 months ago
Permissions
No special permissions declared or detected