You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We are growing a PR backlog given an increase in AI generated code reviews. We generally appreciate fantastic new contributions, though they can take a little longer to review and the increased volume makes it a little harder to keep up.
Poposal
Codify code quality standards to help agents write higher quality code and help with code review. In the future we can also add more conformance tests for common patterns we want to enforce and shift left.
To ensure high-quality contributions and automated reviews across all major coding agents (Google Antigravity, OpenAI Codex, Anthropic Claude Code, and GitHub Copilot for both VS Code and PR Code Review), we can adopt a unified layout that relies on a single source of truth with zero content duplication.
Proposed Minimal File Structure
.
├── AGENTS.md # Single source of truth for guidelines, standards & commands
├── CLAUDE.md -> AGENTS.md # Symlink for Claude Code instructions
├── .agents/
│ └── skills/ # Canonical Agent Skills directory (open specification)
│ └── <skill-name>/
│ └── SKILL.md # Portable Agent Skills spec (progressive disclosure)
└── .claude/
└── skills -> ../.agents/skills # Symlink so Claude Code natively discovers all skills
Activity
allenporter commentedon Sep 12, 2026
Proposal: Unified Agent Instructions & Skills Configuration
To ensure high-quality contributions and automated reviews across all major coding agents (Google Antigravity, OpenAI Codex, Anthropic Claude Code, and GitHub Copilot for both VS Code and PR Code Review), we can adopt a unified layout that relies on a single source of truth with zero content duplication.
Proposed Minimal File Structure
allenporter commentedon Sep 12, 2026
We do not need
copilot-instructions.mdanymore since the only thing that only reads it is VSCode built in review, but normal chat works.AGENTS.md.agents/skills/AGENTS.md.agents/skills/AGENTS.md(noGEMINI.mdrequired) and supports progressive disclosure for skills.AGENTS.md&copilot-instructions.md.agents/skills/AGENTS.md(Chat/Agent) &copilot-instructions.md(Review).agents/skills/.agents/skills/.CLAUDE.md(resolves toAGENTS.md).claude/skills/(resolves to.agents/skills/)/<skill-name>slash commands.docs: add repository engineering guidelines, review skill, and contri…