--- title: "Agent briefing: the context an AI coding agent gets before it starts · Cork" description: "An agent briefing is the targeted context an AI coding agent gets before it works: signatures, callers, recent changes and notes for the code it will touch." canonical: "https://corkcli.com/blog/agent-briefing" last_updated: "2026-11-14" published: "2026-10-03" author: "Kevin Vera" --- # Agent briefing: the context an AI coding agent gets before it starts An agent briefing is the targeted context an AI coding agent gets before it works: signatures, callers, recent changes and notes for the code it will touch. By Kevin Vera · October 3, 2026 · [Glossary](https://corkcli.com/blog/c/glossary) · 3 min read An **agent briefing** is the targeted context handed to an AI coding agent at the start of a session or with a new prompt, describing the code it is about to work on (signatures, callers, recent changes, who else is working there, and decisions already made) so it can act without rediscovering all of it. A briefing is different from general project instructions. Instructions say how to work in this repository. A briefing says what is true right now about the specific code this task touches. ## Why it matters Coding agents start with little memory of the codebase. Claude Code's docs say it directly: "Each Claude Code session begins with a fresh context window." Subagents begin the same way, with "a fresh, isolated context window" that does not include the main conversation's history. So each new agent spends its first steps searching and reading files to rebuild a picture another agent may have built minutes ago. That rediscovery has two costs. It spends tokens and time. And it is often stale: an agent that read a file before a teammate's rename will write code against the old name. ## How it works A briefing has three parts: 1. **Selection.** Decide which code the task is about, from file paths, symbol names or the meaning of the prompt. 2. **Facts about that code.** Exact signatures, the callers that depend on them, recent changes and who made them, relevant tests. 3. **Shared knowledge.** Notes and decisions other agents or people attached to that code, and who is working on it at the moment. It is delivered where the agent will read it without asking: at session start and alongside each prompt. In Claude Code, `SessionStart` and `UserPromptSubmit` are documented hook events, and hook output can add context to the session. ## Examples - Before an agent edits `billing/invoice.py`, it receives the signature of `create_invoice`, its eleven callers, and a note that a teammate changed its currency handling an hour ago. - A prompt says "fix the login redirect". The briefing names the two functions involved, the test that covers them, and that another agent is editing one of them right now. - A new session in a repository gets a short map of the modules it is likely to touch instead of reading a dozen files. ## How teams handle it - **Instruction files.** CLAUDE.md or AGENTS.md hold conventions and architecture notes. They are loaded every session but are not specific to the task and go stale. - **Hand-written task prompts.** A person or lead agent writes the context into the prompt. Claude Code's agent teams docs advise including "task-specific details in the spawn prompt." - **Retrieval.** Search or embeddings pick relevant files to include. - **Code intelligence.** Language servers and code graphs answer "who calls this" on demand. Cork automates briefings. Cork is the [coordination layer](https://corkcli.com/blog/coordination-layer-ai-coding-agents) for multiplayer coding agents: it keeps Claude Code, Codex and Cursor agents, on one machine or across a team's machines, from overwriting each other on one codebase, and hands each agent what the others already learned. When an agent starts or gets a prompt, Cork's hooks hand it the exact signatures, callers and recent changes of the code it is about to touch, with no extra model calls, so it searches less. On the paid plan, prompts are matched to code by meaning, using an embedding model. ## Related terms - [Agent handoff](https://corkcli.com/blog/agent-handoff) - [Blast radius](https://corkcli.com/blog/blast-radius) - Coordination layer (for AI coding agents) - [Multiplayer coding agents](https://corkcli.com/blog/multiplayer-coding-agents) ## Sources - - - - ## Related reading - [Blast radius in code: everything a change can break](https://corkcli.com/blog/blast-radius): The blast radius of a code change is the set of callers, dependents, tests and teammates' work that the change can affect, used to judge risk before editing. - [Agent handoff: passing work and context from one AI agent to another](https://corkcli.com/blog/agent-handoff): An agent handoff is when one AI coding agent passes unfinished work, findings or decisions to another agent or person, so the next one need not start over. - [Blast radius: how to know what a code change breaks before you make it](https://corkcli.com/blog/blast-radius-code-change): Find a code change's blast radius before you make it: callers via LSP and git grep, dependents with Nx or Turborepo, history, tests, and teammates' work. - [Code graphs for AI coding agents: why agents need one](https://corkcli.com/blog/code-graph-for-ai-coding-agents): Why AI coding agents work better with a code graph: fewer file reads, exact callers before an edit, and a map that stays current as several agents edit. [All articles](https://corkcli.com/blog.md) · [Cork in Markdown](https://corkcli.com/index.md) · [llms.txt](https://corkcli.com/llms.txt) This page in HTML: https://corkcli.com/blog/agent-briefing