Point any agent (Claude / Codex / Gemini / a dev-through-an-agent) at this file and it can start developing. This is the onboarding contract for firevector. Keep it current — it's the first thing an incoming agent reads. Human teammates: this doubles as the repo's quickstart.
<One paragraph: what it does, who it's for, why it exists. No jargon.>
<The active goal — the next thing to build. This is the "get rolling" section.>
- Now: <the current objective / the branch>
- Task board / open items: <link or command, e.g.
docs/…-HANDOFF.md, an issue tracker, a board> - Latest context: <link to a handoff/status doc so you don't re-derive anything>
- Branch off
<main>, work in a worktree, open a PR — never commit to<protected>directly. - Never commit secrets. They live in
<where, e.g. ~/.ziggy/secrets.env>, outside the repo. - Do not touch: <control planes / generated files / third-party clones / real user data>.
- Match the surrounding code's style, naming, and comment density.
- Full multi-agent discipline: <link to AGENTS.md or this repo's equivalent>.
<clone / install>
<how to run it locally>
<how to test / how to confirm a change actually works>
- Declare yourself on every change: your agent name + one line on what you're doing, in the commit
trailer (
Co-Authored-By: <you>) and any run log. Honesty about what you did/verified is the rule. - (Emerging) verifiable identity: signing commits with a per-agent keypair and registering the public key is the direction we're heading — start with a declared name; we'll standardize keys as the fleet grows. Don't block on it now.
- Docs index: · Owner / who to ask: <name/contact> · Related repos:
Meta: maturity <early|growing|established> · agent(s) <none|which> · last reviewed <date>.