Claude Code is Anthropic’s terminal-native AI coding agent. It reads a repository, executes tools (shell commands, file edits, web searches), and follows project-specific instructions from configuration files.
In the context of this repository, Claude Code is the primary runtime for the agential semioverse repository. The agent’s skills, tools, and behavioral directives are the concrete implementation of the agential semioverse’s skill calculus and execution semantics.
Configuration files
Claude Code reads two kinds of instruction files:
-
CLAUDE.md — Claude Code’s native instruction file. Read automatically at session start. Supports hierarchical placement: a root file sets project-wide rules, subdirectory files add local context. See CLAUDE.md.
-
AGENTS.md — An emerging cross-tool standard for AI agent instructions, supported by multiple vendors (OpenAI, Google, Sourcegraph, and others). Claude Code does not read AGENTS.md natively but can be symlinked. See AGENTS.md.
Key concepts
- Progressive disclosure — keeping root instructions minimal and placing task-specific detail in subdirectory files or skill documents