scawful/yaze
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

backend-infra-engineer: Release v0.3.9-hotfix7 snapshot

Browse Source
This commit is contained in:
scawful
2025-11-23 13:37:10 -05:00
parent c8289bffda
commit 2934c82b75
202 changed files with 34914 additions and 845 deletions
Download Patch File Download Diff File Expand all files Collapse all files

67
AGENTS.md
View File

@@ -1,45 +1,32 @@
## Inter-Agent Collaboration Protocol
## Inter-Agent Protocol (Lean)
1) **Read the board** (`docs/internal/agents/coordination-board.md`) before starting.
2) **Log your intent** (task, scope, files) on the board with your agent ID.
3) **Answer requests** tagged to your ID.
4) **Record completion/handoff** with a short state note.
5) For multi-day work, use `docs/internal/agents/initiative-template.md` and link it from your board entry.
Multiple assistants may work in this repository at the same time. To avoid conflicts, every agent
must follow the shared protocol defined in
[`docs/internal/agents/coordination-board.md`](docs/internal/agents/coordination-board.md).
## Agent IDs (shared with Oracle-of-Secrets/.claude/agents)
Use these canonical IDs (scopes in `docs/internal/agents/personas.md` and `.claude/agents/*`):
### Required Steps
1. **Read the board** before starting a task to understand active work, blockers, or pending
requests.
2. **Append a new entry** (format described in the coordination board) outlining your intent,
affected files, and any dependencies.
3. **Respond to requests** addressed to your agent ID before taking on new work whenever possible.
4. **Record completion or handoffs** so the next agent has a clear state snapshot.
5. For multi-day initiatives, fill out the template in
[`docs/internal/agents/initiative-template.md`](docs/internal/agents/initiative-template.md) and
link it from your board entry instead of duplicating long notes.
| Agent ID | Focus |
|----------------------------|--------------------------------------------------------|
| `ai-infra-architect` | AI/agent infra, z3ed CLI/TUI, gRPC/network |
| `backend-infra-engineer` | Build/packaging, CMake/toolchains, CI reliability |
| `docs-janitor` | Docs, onboarding, release notes, process hygiene |
| `imgui-frontend-engineer` | ImGui/renderer/UI systems |
| `snes-emulator-expert` | Emulator core (CPU/APU/PPU), perf/debugging |
| `test-infrastructure-expert` | Test harness, CTest/gMock, flake triage |
| `zelda3-hacking-expert` | Gameplay/ROM logic, Zelda3 data model |
| `GEMINI_FLASH_AUTOM` | Gemini automation/CLI/tests |
| `CODEX` | Codex CLI assistant |
| `OTHER` | Define in entry |
### Agent IDs
Use the following canonical identifiers in board entries and handoffs (see
[`docs/internal/agents/personas.md`](docs/internal/agents/personas.md) for details):
Legacy aliases (`CLAUDE_CORE`, `CLAUDE_AIINF`, `CLAUDE_DOCS`) → use `imgui-frontend-engineer`/`snes-emulator-expert`/`zelda3-hacking-expert`, `ai-infra-architect`, and `docs-janitor`.
| Agent ID | Description |
|-----------------|--------------------------------------------------|
| `CLAUDE_CORE` | Claude agent handling general editor/engine work |
| `CLAUDE_AIINF` | Claude agent focused on AI/agent infrastructure |
| `CLAUDE_DOCS` | Claude agent dedicated to docs/product guidance |
| `GEMINI_FLASH_AUTOM` | Gemini agent focused on automation/CLI/test work |
| `CODEX` | This Codex CLI assistant |
| `OTHER` | Any future agent (define in entry) |
## Helper Scripts (keep it short)
Located in `scripts/agents/`:
- `run-gh-workflow.sh`, `smoke-build.sh`, `run-tests.sh`, `test-http-api.sh`
Log command results + workflow URLs on the board for traceability.
If you introduce a new agent persona, add it to the table along with a short description.
### Helper Scripts
Common automation helpers live under [`scripts/agents/`](scripts/agents). Use them whenever possible:
- `run-gh-workflow.sh` – trigger GitHub workflows (`ci.yml`, etc.) with parameters such as `enable_http_api_tests`.
- `smoke-build.sh` – configure/build a preset in place and report how long it took.
- `run-tests.sh` – configure/build a preset and run `ctest` (`scripts/agents/run-tests.sh mac-dbg --output-on-failure`).
- `test-http-api.sh` – poll the `/api/v1/health` endpoint once the HTTP server is running.
Log command results and workflow URLs on the coordination board so other agents know what ran and where to find artifacts.
### Escalation
If two agents need the same subsystem concurrently, negotiate via the board using the
`REQUEST`/`BLOCKER` keywords. When in doubt, prefer smaller, well-defined handoffs instead of broad
claims over directories.
## Escalation
If overlapping on a subsystem, post `REQUEST`/`BLOCKER` on the board and coordinate; prefer small, well-defined handoffs.