Project overview in Memory

An AI agent reads your repository and writes a project overview in your SideHub Memory. Nothing to set up, no secret, no extra tool.

The fastest way to see a workflow run on your own code. An agent running on your machine reads the repository and writes the page every newcomer wishes existed: what the project does, how it is built, how to run it and where to start.

How it works

1. The agent explores the repository: README, conventions files (CLAUDE.md, AGENTS.md, CONTRIBUTING.md), manifests, directory structure, entry points and tests.
2. It writes a "Project overview" page in your SideHub Memory: purpose, stack, architecture, how to build, run and test, conventions, and a few good first tasks.
3. Run it again later: it reads the previous overview and writes an up-to-date one.

What you need

• An agent connected to the repository's machine, with Claude Code, Codex or Antigravity logged in. That's all: no secret, no GitHub token, no extra tool. The agent only reads the code, it changes nothing.

Steps

1. 1. Write the project overview

   Write an overview of the git repository in the current directory, for a developer who has never seen it. Read
   only: change no file of the repository, install nothing.
   
   1. If a previous overview exists in the Drive (`sidehub-cli drive search "Project overview"`), read it: keep what
      is still true, update the rest.
   2. Read what the repository says about itself: README, CLAUDE.md, AGENTS.md, CONTRIBUTING.md and docs/ if present.
   3. Read the manifests (package.json, *.csproj, pyproject.toml, go.mod, Cargo.toml, pom.xml…), the build and CI
      configuration, the directory structure, the entry points and where the tests live.
   4. Write the overview in Markdown, with these sections:
      - **Purpose**: what the project does and for whom, in three lines.
      - **Stack**: languages, frameworks, key dependencies and their role.
      - **Architecture**: the main parts, how they talk to each other, the patterns used, with the directories involved.
      - **Build, run and test**: the exact commands, as the repository defines them.
      - **Conventions**: code style, branches, commits, anything the repository asks contributors to follow.
      - **Good first tasks**: three to five small, concrete improvements you noticed (missing tests, unclear docs,
        outdated dependencies), each with the file involved.
   Only state what you read in the repository; when something is unclear, say so instead of guessing.

FAQ

Does it change my code?

No. The agent only reads the repository; the one thing it writes is the overview page in your SideHub Memory.

Where does the code go?

Nowhere. The agent runs on your own machine or server; SideHub only sees the run's status and the page it writes.

Which CLI does it run with?

The one you pick at launch, among those installed on the agent's machine (Claude Code, Codex or Antigravity).

Use this template (free) · All templates