Instruction file imported from avadhootdivekar/agent-orchestrator (.cursor/rules/agent-orchestrator.mdc). Copyright stays with the author.

agent-orchestrator — Cursor rules (pointer)

Name: agent-orchestrator
Author: avadhootdivekar

The authoritative, hand-maintained agent instructions are Claude-native. Do not duplicate them here — read them:

Project context, design principles, rules, commands → @CLAUDE.md
Specialized role guides → .claude/agents/ (architect, manager, dev-epic, developer, reviewer, tester, dev-security, dev-critic)
Skills → .claude/skills/ · Command workflows → .claude/commands/
Ticket workspace (epics/tasks + conventions) → meta/tickets/ (meta/tickets/README.md)
Cross-tool overview → @AGENTS.md

Project in one line

A Python agent-orchestration framework: drives multi-agent workflows to completion via a DAG of tasks (inputs/outputs as artifacts/file paths, cron + event triggers), fully config-driven by structured JSON/YAML spec files that reference payloads (md/source/dirs) by path.

Working rules (summary — CLAUDE.md is authoritative)

Smallest correct change; reuse existing patterns before new abstractions; no magic literals.
Validate structured specs against their schema; reference payloads by path, never inline them into the engine.
Keep core logic deterministic and resumable — inject RNG/clock; tasks idempotent and safe to retry.
Tests with pytest; lint/format/types via ruff + mypy. Never claim tests pass without running them.
When a task needs a specialized role, open the matching .claude/agents/<role>.md and follow it.

Edit the .claude/ originals and CLAUDE.md, not this pointer.

{ "ocm": "1", "id": "avadhootdivekar-agent-orchestrator-agent-orchestrator-rule", "kind": "skill", "name": "agent-orchestrator", "description": "agent-orchestrator project rules — pointer to the authoritative Claude-native instructions.", "publisher": "avadhootdivekar", "version": "1.0.0", "capabilities": { "domains": [ "general" ], "tags": [ "skill-md", "github-cursor-rules" ], "languages": [ "en" ] }, "quality_prior": 0.6, "examples": [ "agent-orchestrator project rules — pointer to the authoritative Claude-native instructions." ], "primary": false, "metadata": { "source": { "provider": "github-cursor-rules", "repository": "https://github.com/avadhootdivekar/agent-orchestrator", "path": ".cursor/rules/agent-orchestrator.mdc", "ref": "b849b7cb95b3aabe1155857334632e5388c35be1", "url": "https://github.com/avadhootdivekar/agent-orchestrator/blob/b849b7cb95b3aabe1155857334632e5388c35be1/.cursor/rules/agent-orchestrator.mdc", "key": "avadhootdivekar/agent-orchestrator/.cursor/rules/agent-orchestrator.mdc" } }, "instructions": "# agent-orchestrator — Cursor rules (pointer)\n\nThe authoritative, hand-maintained agent instructions are Claude-native. **Do not duplicate them here — read them:**\n\n- Project context, design principles, rules, commands → @CLAUDE.md\n- Specialized role guides → `.claude/agents/` (`architect`, `manager`, `dev-epic`, `developer`, `reviewer`, `tester`, `dev-security`, `dev-critic`)\n- Skills → `.claude/skills/` · Command workflows → `.claude/commands/`\n- Ticket workspace (epics/tasks + conventions) → `meta/tickets/` (`meta/tickets/README.md`)\n- Cross-tool overview → @AGENTS.md\n\n## Project in one lin", "cost": { "context_tokens": 382 } }

agent-orchestrator

About

agent-orchestrator — Cursor rules (pointer)

Project in one line

Working rules (summary — CLAUDE.md is authoritative)

Use it

Manifest

Reviews