Skip to content

internal-dogfood.md

Latest commit

 

History

History
98 lines (71 loc) · 3.91 KB

internal-dogfood.md

File metadata and controls

98 lines (71 loc) · 3.91 KB
Raw

Internal Dogfood Wiring

Brigade is repo-wired first and machine-aware second.

The repo wiring lives under .brigade/ and is intentionally gitignored. It stores local config, receipts, scans, reports, queues, readiness waivers, and run artifacts for one repo. The machine wiring is limited to installed commands and user-level optional tools such as brigade, codex, and agent-notify.

Internal dogfood is repo wiring plus an explicit daily loop

Onboard a Repo

Use the internal dogfood profile when a repo should participate in the local production loop:

brigade operator init --profile internal-dogfood --target .
brigade operator sync-tools --target .
brigade operator doctor --profile internal-dogfood --target .
brigade operator status --profile internal-dogfood --target .

For Brigade's own repo, public release readiness can be waived when the goal is internal production rather than publishing:

brigade operator init --profile internal-dogfood --target . --waive-public-release

That waiver is local evidence only. It does not publish, tag, push, mutate remotes, or change the public release gate.

Daily Agent Loop

Agents and humans should use Brigade as an explicit local loop, not as a daemon:

brigade operator guide
brigade operator doctor --profile internal-dogfood --target .
brigade operator status --profile internal-dogfood --target .
brigade operator sync-tools --target .
brigade daily status --target .
brigade daily plan --target .

When the selected action is a safe bounded adapter, run:

brigade daily run --target .

When the selected action needs judgment or external setup, do it manually, then run brigade daily status --target . again and confirm the signal moved.

Handoffs

If a session changes how Brigade should be used in this repo, write that durable knowledge into the normal handoff path for the active harness:

  • Codex dogfood runs default to .codex/memory-handoffs/.
  • Claude Code uses .claude/memory-handoffs/.
  • OpenCode uses .opencode/memory-handoffs/.
  • Antigravity uses .antigravity/memory-handoffs/.
  • Pi uses .pi/memory-handoffs/.
  • Cursor uses .cursor/memory-handoffs/.
  • Aider uses .aider/memory-handoffs/.
  • Goose uses .goose/memory-handoffs/.
  • Continue uses .continue/memory-handoffs/.
  • GitHub Copilot CLI uses .copilot/memory-handoffs/.
  • Qwen Code uses .qwen/memory-handoffs/.
  • Kimi Code uses .kimi/memory-handoffs/.
  • AdaL uses .adal/memory-handoffs/.
  • OpenHands uses .openhands/memory-handoffs/.
  • Grok CLI uses .grok/memory-handoffs/.
  • Hermes uses .hermes/memory-handoffs/.

Prefer the writer command so the draft matches Brigade's expected section style and lints before handoff:

brigade handoff draft --title "Brigade workflow update" --summary "What changed and why it matters." --content "### Brigade workflow update\n\nDurable guidance to append."

Handoffs should mention the concrete commands, what changed, what remains manual, and any local-only waiver or setup assumption. Do not copy raw .brigade/ evidence into committed docs.

Boundaries

The internal dogfood profile:

  • writes repo-local gitignored config defaults
  • refreshes read-only security evidence
  • reports repo-local versus machine-level wiring
  • projects tracked tools/*.md sources into local harness folders when brigade operator sync-tools --target . is run
  • verifies a harness handoff inbox with brigade operator verify-harness --harness hermes --target .
  • keeps public release readiness separate from internal production readiness

It does not:

  • start background services
  • install global hooks
  • send notifications
  • ingest handoffs into canonical memory
  • publish, push, tag, or mutate remotes
  • make other repos participate automatically

Other repos need their own brigade operator init --profile internal-dogfood --target <repo> or need to be tracked through the repo-fleet commands.