hebel. today · 14 November 202624 / 80

Agentic Engineering · Context Engineering

Engineers pivot from prompt engineering to structured component manifests

Design and product teams are shifting from prompt engineering to creating machine-readable manifests that define component properties, variants, and defaults explicitly. By generating these inventories directly from code, teams ensure agents pick from a…

Design and product teams are shifting from prompt engineering to creating machine-readable manifests that define component properties, variants, and defaults explicitly (UX Collective). By generating these inventories directly from code, teams ensure agents pick from a closed list rather than inventing styles or variants that do not exist in the production design system (UX Collective). This approach uses the Model Context Protocol to allow agents to query the current state of a system in real time, reducing the reliance on stale, memory-based assumptions. Beyond static data, senior engineers at Google DeepMind architect agents to treat decision context as a first-class input, pairing critiques with concrete, feasibility-aware prescriptions that account for specific project deadlines (Hao Hoang from AI Interview Prep).

  • William Imoh uses a team of distinct Grok Bots, each with a clear persona and specific role, to manage business operations and delegate tasks effectively.
  • Elastic suggests pre-computing enterprise data into an AI index to reduce token burn and prevent agents from accessing stale information during complex tasks (Alex Wilhelm from The New Stack).
  • Jeremy Wright maintains a single convention for handoff files across projects to capture the reasoning behind decisions and prevent agents from suggesting rejected options (The ECHO Files).
  • Porchlight stores company-wide context in a hex.md file, ensuring the AI analyst understands business conventions and recurring mistakes without needing irrelevant domain-specific guidance (Shane Butler from AI Analyst Lab).

Try today · 30 min

Generate a JSON manifest of your primary design system components.

Run a script over your component library to extract names, properties, and allowed values.

You need

  • Access to your component source code
  • A script or tool to parse component types
  • A JSON file template

Steps

  1. Identify the core components you use most frequently.
  2. Extract the name, variant, size, and property definitions for each.
  3. Format these attributes into a structured JSON list.
  4. Verify the output matches the actual production component types.
  5. Save the file as manifest.json in your project root.

One email instead of your newsletter pile.

Forward the newsletters you already get (a Gmail filter does it) and tell Hebel your goals. Every morning it reads them with 20,000 others and sends one short email, with one small thing to do today. Fewer emails, not more. Free during beta.

Start my paper →
LinkedIn X

Sources

  1. UX Collective on hebel. →This guide details how to extract component props and metadata into a machine-readable manifest for AI prototyping.

  2. Hao Hoang, AI Interview Prep on hebel. → ✎ Written by a personThis interview outlines the architecture for context-aware agents that prioritize feasibility and specific user decision constraints.

  3. Alex Wilhelm, The New Stack on hebel. → ✎ Written by a personThis article describes how pre-computing enterprise metadata into an AI index lowers token costs and improves response accuracy.

  4. The ECHO Files on hebel. →This post emphasizes the need for handoff files that record the reasoning behind decisions to prevent agents from revisiting failed paths.

  5. Shane Butler, AI Analyst Lab on hebel. → ✎ Written by a personThis tutorial shows how to structure company-wide context using dedicated markdown files to align AI analysts with business conventions.

More in Agentic Engineering

Autonomous agent reliability hinges on orchestration versus human oversight

Building reliable agent harnesses

Production agents stay safe with rigid action catalogs

LinkedIn automates 40,000 weekly pull requests

Context Engineering yesterday

Anthropic and agents use 'dreaming' to curate memory

Sign in to Hebel first

Likes, your collection and done Hebel need an account. We email you a six digit code; no password.