Context Knowledge — Development skill for Claude Code

context-knowledge

**A context harness for Claude Code project understanding and safe changes**

`context-knowledge` is a Claude Code Skill for building and maintaining a structured project context knowledge base before an AI agent changes an unfamiliar codebase.

Why it exists

When an agent enters a project for the first time, it usually does not know the project structure, module responsibilities, file dependencies, existing constraints, or the difference between implemented behavior and future design goals. Without that context, an agent may:

• edit the wrong file or module;
• overlook dependencies between project areas;
• mistake a planned feature for an implemented capability;
• violate existing security, API, data, or runtime constraints;
• expand the requested scope without confirmation;
• write, commit, publish, or otherwise act before the user understands the proposed change.

`context-knowledge` addresses this by creating a structured summary of the project that the agent can verify against source code and authoritative documents before making changes. It is a compact, source-linked model of the project—not a replacement for the source code.

What it helps with

Use this Skill to:

• build a reliable first-pass understanding of an unfamiliar project;
• document module boundaries, data flow, dependencies, and external-system boundaries;
• distinguish implemented facts, project constraints, design goals, and unresolved items;
• identify the files and systems affected by a proposed change;
• surface risks, conflicts, validation requirements, and knowledge-base impact;
• require explicit user confirmation before modifying files;
• reduce accidental edits, scope expansion, and changes made in the wrong place.

It is intended for new features, refactors, bug fixes, architecture changes, and database, API, UI, configuration, or external-service changes.

What it is

`context-knowledge` is a **workflow and context-governance Skill**, also described