It would be great to narrow down to a standard as contributors to the same repo can use different agents:
| Coding Agent / Tool |
Directory for Agent Specs / Prompts |
Notes |
| Kiro |
.kiro/specs |
Native Kiro spec location |
| Cursor |
.cursor/rules
.cursor/prompts |
Rules + reusable prompts |
| GitHub Copilot |
.github/copilot-instructions.md
.github/prompts/ |
Repo-level Copilot guidance |
| Claude (Anthropic) |
.claude/instructions.md
.claude/prompts/ |
Claude Code conventions |
| OpenAI / ChatGPT |
.chatgpt/instructions.md
.chatgpt/prompts/ |
Repo-scoped instructions |
| Aider |
.aider.conf.yml
.aider/prompts.md |
Config + prompt file |
| Continue.dev |
.continue/config.json
.continue/prompts/ |
JSON config + prompt dir |
| Sweep AI |
.sweep/config.yaml
.sweep/prompts/ |
Automation-focused agent |
| Sourcegraph Cody |
.sourcegraph/cody.json
.sourcegraph/prompts/ |
Enterprise-friendly setup |
| Tabnine |
.tabnine/config.json |
Limited prompt support |
| Auto-GPT / Agent Frameworks |
agents/
agents/specs/
prompts/ |
Flexible, framework-defined |
| LangChain / CrewAI |
agents/
agents/specs/
prompts/ |
Common open-agent pattern |
Originally posted by @kmturley in agentsmd/agents.md#71 (comment)