Glossary

Coding agent

An AI program that does programming tasks on its own. Given a written request, it reads the codebase, edits files, runs commands and tests, and keeps going until the task is done. Claude Code, Codex, Cursor and Gemini CLI are examples. It carries out a whole task rather than suggesting the next line.

Instruction file CLAUDE.md, AGENTS.md

A plain text file in a codebase that tells a coding agent how the team works: conventions to follow, commands to run, things never to do. It's the agent's handbook. CLAUDE.md is the instruction file Claude Code reads; AGENTS.md is a shared format that Codex and other agents read. The agent reads it at the start of every session, so editing it changes what the agent does.

Rules

Smaller instruction files for one part of the codebase, such as “how we write payment code”. Some load only when the agent works on matching files, so if files move or a pattern is mistyped, a rule can stop loading without anyone noticing. In Claude Code they live in .claude/rules/.

Skills

Packaged instructions for one kind of job, like releasing a new version. The agent always sees a short description of each skill and reads the full skill only when it decides it needs it, so the description decides when the skill gets used.

MCP tool

A tool the agent can use beyond the codebase, such as a database, a ticket tracker or a documentation search. MCP (Model Context Protocol) is a standard way to plug such tools into AI agents. Each tool comes with a description the agent reads to decide when and how to use it, so rewording that description can change the agent's behavior.

Model upgrade

Switching the AI model behind the agent to a newer or different one. It changes everything at once: some tasks get better, others get worse, and instructions written for the old model may work differently with the new one.

Context

Everything the agent reads before and while it works: instruction files, rules, skills, tool descriptions and settings. A “context diff” on this site is the list of what the agent will see differently after a change.

Pull request

A proposed change to a codebase, shown as the lines added and removed, that teammates review before it's accepted (“merged”). Changes to an agent's instruction files go through pull requests too. GitLab calls them merge requests.

Regression

Something that used to work and got worse after a change. On this site, a task the agent used to complete reliably that it now completes less often. Rerunner will report it as REGRESSED: worse.

CI

Continuous integration: the automatic checks that run on every pull request, such as building the code and running its tests, before anyone merges it. A failing check can block the merge. Rerunner is designed to be one of those checks, for the agent instead of the code.

Read the problem, in plain words