liveBook · Manning

Context Fabric

As explained in Vibe Engineering

Context Fabric is a persistent layer of specifications, conventions, documentation, and structured knowledge surrounding a software project that supplies agents with project-specific working context.

An agent may see the files left open but miss the dependency graph, project conventions, and the developer’s intent. A Context Fabric addresses that gap: it is a persistent layer of specifications, documentation, conventions, and structured knowledge that makes project context available across agent sessions. section 3.0 section 3.7

The fabric treats documentation as infrastructure. Missing context can cost tokens, extra iterations, and wrong code; the person who skipped the documentation may pay that cost almost immediately. The layer must also be maintained and curated: giving an agent more context is not automatically better. section 3.5.4 section 3.3.6 section 3.7

A practical fabric can include:

  • Specs, memory files, and structured project knowledge that persist outside a single session. section 3.7
  • Curated integrations and approved tools that agents can use without each developer assembling them again. section 3.3.6
  • Versioned procedures, such as review checklists and deploy scripts, packaged for the team to install and reuse. section 3.3.6
Listing 3.1 Skills.md
---
name: pull-request-review
description: Reviews a pull request before it reaches a human
  reviewer. Use when asked to review, check, or sanity-check a
  PR or a branch diff.
---
# Pull request review
1. Run scripts/check-migrations.sh. If it exits non-zero, stop
   and report - do not review further.
2. Walk the diff against references/review-checklist.md.
3. Report findings as blocking, non-blocking, or nitpick.
Never approve a PR. A human approves.

This pull-request review skill packages a repeatable procedure: stop if the migration check fails, classify findings, and leave approval to a human. section 3.3.6

A fabric does not replace direction or supervision. Define the goal, constraints, and acceptance criteria; then inspect the agent’s context when it goes wrong. For example, redirect it from outdated documentation to the relevant migration guide rather than assuming retrieval selected the right source. section 3.5.4

How Vibe Engineering covers it

How deep it goes
Defines the persistent layer and explains how curated integrations and versioned procedures make it transferable across a team.
Examples
Uses a pull-request review skill with a migration check, review checklist, findings categories, and human approval.
What you can do after
Diagnose/debug agent context by correcting irrelevant retrieval and directing the agent to the appropriate documentation.
How it connects
Connects context engineering with retrieval-augmented generation, context compaction, and MCP integrations.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Vibe Engineering. The section links point to the parts of the book it draws on.