Imported from thomHayner/scaleforce-github-orchestrator (AGENTS.md). Install upstream with npx skills add thomHayner/scaleforce-github-orchestrator. Copyright stays with the author.

AGENTS.md

Conventions for all AI agents working in this repo (Claude, Copilot, Codex, Cursor, etc.). Agent-specific rules live in CLAUDE.md, .github/copilot-instructions.md, etc.

Who's who

Agents have distinct identities on purpose. Do not impersonate the maintainer or another agent.

Universal rules

1. Preserve commit attribution. Commits are authored by the maintainer's git identity; AI contribution is recorded via Co-Authored-By: trailers. Never modify git config user.*.
2. Don't @-mention the maintainer when posting under their identity. It self-tags them.
3. Architecture-affecting changes require an ADR in docs/adr/ (MADR format). If you're not sure whether your change qualifies, err on the side of writing one.
4. Branch model: feature/* → dev → main. main is release-only. See docs/setup/branching.md.
5. One logical change per PR. Refactors separate from features.
6. Read docs/ before starting nontrivial work: vision → roadmap → specs → adr.
7. Update llm-wiki/ when you learn something durable that future agents will need (a non-obvious convention, an integration quirk, a decision rationale not yet in an ADR).

Where to find things

• Product context: docs/
• Synthesized agent-readable knowledge: llm-wiki/
• Repo automations: .github/workflows/
• Issue/PR/Discussion templates: .github/
• AI-internal label templates (used by scaleforce[bot] to apply labels): .github/UserNoAiYesTemplates/

Per-agent guides

• Claude → CLAUDE.md
• Copilot → .github/copilot-instructions.md (create when needed)
• Codex / Cursor → read this file plus CLAUDE.md; the rules are the same.