{"id":48832165,"url":"https://github.com/maxtechera/orchestrator","last_synced_at":"2026-04-14T21:01:00.928Z","repository":{"id":348994644,"uuid":"1178500361","full_name":"maxtechera/orchestrator","owner":"maxtechera","description":"Skill-first orchestration framework. Dispatch agents, verify deliverables, improve with every cycle.","archived":false,"fork":false,"pushed_at":"2026-04-13T19:59:45.000Z","size":158,"stargazers_count":1,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-04-13T20:06:54.380Z","etag":null,"topics":["ai-agents","autonomous","claude-code","openclaw","orchestration","skill-first","verification"],"latest_commit_sha":null,"homepage":null,"language":null,"has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/maxtechera.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":"CODE_OF_CONDUCT.md","threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":"SECURITY.md","support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-03-11T04:41:32.000Z","updated_at":"2026-04-13T19:59:48.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/maxtechera/orchestrator","commit_stats":null,"previous_names":["maxtechera/orchestrator"],"tags_count":1,"template":false,"template_full_name":null,"purl":"pkg:github/maxtechera/orchestrator","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/maxtechera%2Forchestrator","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/maxtechera%2Forchestrator/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/maxtechera%2Forchestrator/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/maxtechera%2Forchestrator/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/maxtechera","download_url":"https://codeload.github.com/maxtechera/orchestrator/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/maxtechera%2Forchestrator/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":31815080,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-14T18:05:02.291Z","status":"ssl_error","status_checked_at":"2026-04-14T18:05:01.765Z","response_time":153,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["ai-agents","autonomous","claude-code","openclaw","orchestration","skill-first","verification"],"created_at":"2026-04-14T21:00:59.858Z","updated_at":"2026-04-14T21:01:00.918Z","avatar_url":"https://github.com/maxtechera.png","language":null,"funding_links":[],"categories":[],"sub_categories":[],"readme":"# /orchestrator\n\n[![MIT License](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)\n[![Version](https://img.shields.io/badge/version-1.0.0-green.svg)](CHANGELOG.md)\n\n**Expert orchestrator of agentic teams. The ticket contract defines work. The verification harness rules completion.**\n\nClaude Code:\n```\n/plugin marketplace add maxtechera/orchestrator\n```\n\nOpenClaw:\n```\nclawhub install orchestrator\n```\n\nZero config. Run `/orchestrator sweep` immediately — if no board is connected, the setup wizard runs automatically.\n\n---\n\n## Install\n\n### Claude Code\n\n#### Install\n```\n/plugin marketplace add maxtechera/orchestrator\n```\n\n#### Update\n```\nclaude plugin update orchestrator@orchestrator\n```\n\n### OpenClaw\n```bash\nclawhub install orchestrator\n```\n\n### Manual\n```bash\ngit clone https://github.com/maxtechera/orchestrator.git ~/.claude/skills/orchestrator\n```\n\n**The agent that did the work never grades its own homework.** A separate verification pass runs with fresh context — real API calls, screenshots, link checks, compliance validation. Not another AI saying \"looks good.\"\n\n**The core pattern:** Ticket contract (Inputs / Deliverables / Verification / Artifacts) → dispatch worker agent → independent verifier → evidence posted to ticket. One cycle, every ticket.\n\n**Best for:** operators running agentic teams who want verified output, not just agent output.\n\nThat's it. Run `/orchestrator sweep` to get started. If no board is connected, the orchestrator will prompt you to run `/orchestrator setup`. See [Your First Ticket](#your-first-ticket) below for a copy-paste template.\n\n---\n\n## What people use it for\n\n**Running a content operation.** `/orchestrator sweep` reads all open tickets, dispatches content agents loaded with your brand voice, and independently verifies every deliverable before marking it done. Not just \"here's a draft\" — checked output.\n\n**Managing an e-commerce store.** Set `BOARD_PROVIDER=linear` and connect Shopify. The orchestrator processes inventory updates, price changes, and product listings — and verifies prices, compare-at fields, and image counts before closing tickets.\n\n**Engineering ticket flow.** `/orchestrator sweep` picks up GitHub Issues, runs domain-specific verification (link checks, schema validation, test presence), and only marks tickets done when the deliverable actually passes.\n\n**Auditing before a release.** `/orchestrator review \u003cticket\u003e` independently checks a specific deliverable. The agent that did the work never grades its own homework.\n\n---\n\n## Setup: Progressive Integration Unlocking\n\nStart using `/orchestrator` immediately. Add integrations when you want more domains.\n\n### 1. Zero Config (built-in skills) — Just install\n\nRun `/orchestrator sweep`. The orchestrator uses your existing board connections (MCP servers for Linear, GitHub, etc.) and built-in agent capabilities. No API keys, no configuration. If no board is detected, you'll be prompted to run `/orchestrator setup`.\n\n### 2. Run the setup wizard (connect your board)\n\n```\n/orchestrator setup\n```\n\nThe setup wizard detects your environment, connects your issue tracker, and validates integrations. Takes about 30 seconds. Supports Linear, GitHub Issues, Notion, and Jira. It writes `BOARD_PROVIDER` and your API key to `~/.config/orchestrator/.env`.\n\nIf you prefer manual config:\n\n```bash\nmkdir -p ~/.config/orchestrator \u0026\u0026 chmod 700 ~/.config/orchestrator\n```\n\nThen add to `~/.config/orchestrator/.env`:\n\n```bash\n# Add to ~/.config/orchestrator/.env\nBOARD_PROVIDER=linear        # or: github, notion, jira\nLINEAR_API_KEY=lin_api_xxxxx  # swap for your tracker's key\n```\n\n### 3. Add domain skills (RECOMMENDED — the single most impactful upgrade)\n\n**Domain skills give agents real expertise** — domain-specific rules, verification criteria, and best practices learned from actual executions. Without them, agents use general capabilities. With them, they follow rules like \"carousel images must be 1080x1350\" and \"always verify compare-at price is cleared when not on sale.\"\n\nDomain skills are **composition wrappers** — they load [`maxtechera/ship`](https://github.com/maxtechera/ship) and [`maxtechera/memory`](https://github.com/maxtechera/memory) skills and add domain-specific verification rules on top. Install ship and memory first, then the domain skill activates their capabilities for orchestrator verification. See [skills/README.md](skills/README.md) for the full list and their composition chains.\n\n```\n/plugin marketplace add maxtechera/skill-content\n/plugin marketplace add maxtechera/skill-ecommerce\n/plugin marketplace add maxtechera/skill-seo\n```\n\n### 4. Add Shopify (e-commerce verification)\n\nConnect your Shopify store for automated price, inventory, and product verification.\n\n```bash\n# Add to ~/.config/orchestrator/.env\nSHOPIFY_ACCESS_TOKEN=shpat_xxxxx\n```\n\n### 5. Add MailerLite (content/newsletter verification)\n\nFree tier at [mailerlite.com](https://mailerlite.com). Enables email sequence verification — link checks, segment validation, send tests.\n\n```bash\n# Add to ~/.config/orchestrator/.env\nMAILERLITE_API_KEY=xxxxx\n```\n\n### 6. Add Apollo (sales verification)\n\nFree tier at [apollo.io](https://apollo.io). Enables lead sourcing verification — bounce rate, compliance, sequence validation.\n\n```bash\n# Add to ~/.config/orchestrator/.env\nAPOLLO_API_KEY=xxxxx\n```\n\n### 7. Optional integrations\n\n```bash\n# Add to ~/.config/orchestrator/.env\nMETA_ADS_TOKEN=xxxxx          # Meta Ads — campaign management\nGOOGLE_ADS_TOKEN=xxxxx        # Google Ads — campaign management\nGA4_PROPERTY_ID=xxxxx         # Google Analytics — conversions\nHUBSPOT_API_KEY=xxxxx         # HubSpot — CRM, pipeline\nSTRIPE_SECRET_KEY=xxxxx       # Stripe — payment verification\nQUICKBOOKS_TOKEN=xxxxx        # QuickBooks — invoicing, AR\nXERO_TOKEN=xxxxx              # Xero — alternative to QuickBooks\nZENDESK_TOKEN=xxxxx           # Zendesk — support ticket management\nMERCADOLIBRE_TOKEN=xxxxx      # MercadoLibre — LATAM marketplace\nINSTAGRAM_ACCESS_TOKEN=xxxxx  # Instagram Graph API — post publishing\n```\n\nThe orchestrator validates every connection before dispatching. Missing integrations fail early with a clear message — not midway through execution.\n\n---\n\n### Do I need API keys?\n\n| Integration | Free? | Do you need it? |\n|-------------|-------|-----------------|\n| Issue tracker | Free | **Yes — pick one.** Linear, GitHub Issues, Notion, or Jira. |\n| Shopify | Paid | **Yes, for e-commerce.** Product pages, pricing, inventory. |\n| MailerLite | Free tier | **Yes, for content.** Email sequences, newsletters. |\n| Apollo | Free tier | **Yes, for sales.** Lead sourcing, outreach. |\n| Instagram | Free (Graph API) | **Optional.** Post publishing, carousels. |\n| Meta Ads | Paid | **Optional.** Campaign management. |\n| Google Ads | Paid | **Optional.** Campaign management. |\n| HubSpot | Free tier | **Optional.** CRM, pipeline. |\n| GA4 | Free | **Optional.** Analytics, conversions. |\n| Stripe | Paid | **Optional.** Payment verification. |\n| QuickBooks | Paid | **Optional.** Invoicing, AR. |\n| Xero | Paid | **Optional.** Alternative to QuickBooks. |\n| Zendesk | Paid | **Optional.** Ticket management. |\n| MercadoLibre | Free | **Optional.** LATAM marketplace listings. |\n| GitHub | Free | **For engineering.** PRs, CI, deployments. |\n\n*All credentials stored locally in `~/.config/orchestrator/.env`. Never transmitted. The orchestrator detects available MCP servers first and uses them when available — the .env is a fallback.*\n\n*The orchestrator receives no money from any integration provider — no referrals, no kickbacks.*\n\n---\n\n### Config file locations\n\n```bash\n# Global config\n~/.config/orchestrator/.env       # API keys and credentials (chmod 600 recommended)\n~/.orchestrator/skills/           # Installed domain skill files\n~/.orchestrator/rules/            # Accumulated operational rules (grow automatically)\n~/.orchestrator/history/          # Outcome logs for self-improvement\n```\n\n### Other platforms\n\nThe Manual install works for Codex CLI, Gemini CLI, and OpenCode — clone to `~/.agents/skills/orchestrator` and the skill is discovered via `SKILL.md`. Add credentials to `~/.config/orchestrator/.env`.\n\n---\n\n## Usage\n\n```\n/orchestrator sweep                            Process all actionable tickets\n/orchestrator sweep TICKET-046                 Re-run a single ticket\n/orchestrator status                           Current ticket states and verification results\n/orchestrator review                           Tickets awaiting your review (pre-verified)\n/orchestrator approve TICKET-044               Approve → Done\n/orchestrator approve TICKET-044 --note \"...\"  Approve with a note\n/orchestrator reject TICKET-044 --note \"...\"   Reject → back to execution with your feedback\n/orchestrator approve-rule \"...\"               Approve a proposed rule into the skill\n/orchestrator reject-rule \"...\"                Reject a proposed rule\n/orchestrator rules --pending                  List proposed rules awaiting approval\n/orchestrator flag TICKET-XXX \"description\"    Report a verification false negative\n/orchestrator setup                            Connect your board and validate integrations\n/orchestrator stats                            This week's pass/fail rate, ticket count, cost\n```\n\n**Reviewing verified work:** When tickets land in Review, run `/orchestrator review` to see the verification report — what passed, what failed, and the evidence. Then approve, reject with a note, or approve with a note. This is your only daily touchpoint.\n\n**When rules are proposed:** After repeated failures with the same root cause, the system proposes a new rule in the review output and via `/orchestrator rules --pending`. Approve or reject inline. Approved rules are added to the skill file and apply to all future tickets.\n\n---\n\n## How It Works\n\nThree steps, every ticket:\n\n1. **Dispatch** — read the board, prioritize by proximity to done (Review \u003e Verification \u003e In Progress stale \u003e Todo), match to domain skill, validate integrations, dispatch worker agent\n2. **Verify** — separate verification pass with fresh context. Automated checks (API assertions, link validation, test suites) + AI judgment (brand voice, content quality). The verifier never knows how the work was done — no confirmation bias.\n3. **Proof** — verification report posted to ticket with evidence. PASS → Done. PARTIAL → Review (human decides). FAIL → auto-retry with failure context, or escalates to Review.\n\nThe harness is the invariant. The worker changes. The verification contract never does.\n\n---\n\n## Your First Ticket\n\nAdd the four sections as markdown headers (`## Inputs`, `## Deliverables`, `## Verification`, `## Artifacts`) in your ticket description. Copy this to your board and run `/orchestrator sweep`:\n\n**Publish SEO blog post**\n\n```\n## Inputs\nKeyword brief \"AI agent verification\", brand voice guide, 3 competitor URLs\n\n## Deliverables\nBlog post published on CMS, meta tags set, internal links added\n\n## Verification\nPage returns 200, word count \u003e 1,500, SEO score green, no brand violations\n\n## Artifacts\nPublished URL, screenshot of live page\n```\n\n**What to expect on your first sweep:**\n- Tickets with all four sections get dispatched\n- Tickets missing sections get moved to Backlog with a comment explaining what's needed\n- If none of your tickets are in this format, the orchestrator tells you and offers to help reformat them\n\nMore templates for every domain in [docs/VISION.md](docs/VISION.md).\n\n---\n\n## Example: What a Verification Report Looks Like\n\nAfter the orchestrator verifies a ticket, this is posted to the ticket:\n\n```\n## Verification Report — TICKET-043\n\nStatus: PASS ✓\n\nChecks:\n  ✓ Page returns 200 (automated — 0.3s)\n  ✓ Word count: 1,847 \u003e 1,500 (automated — 0.1s)\n  ✓ SEO score: 92/100, green (automated — 1.2s)\n  ✓ Meta title set, under 60 chars (automated — 0.1s)\n  ✓ og:image present and resolves (automated — 0.4s)\n  ✓ 3 internal links found (automated — 0.2s)\n  ✓ Brand voice check: consistent, no violations (AI judgment — 2.1s)\n\nEvidence:\n  - Screenshot: live page captured\n  - Lighthouse report: attached\n  - All links validated: 3/3 resolve\n\nResult: PASS — moved to Done\n```\n\n---\n\n## Domain Skills\n\nDomain skills extend the orchestrator with specialized verification rules for specific work types. Each skill is a composition wrapper — it adds domain rules and verification criteria on top of the base harness.\n\n| Skill | What it adds | Plugin |\n|-------|-------------|--------|\n| content | Brand voice, format specs, engagement baseline | `maxtechera/skill-content` |\n| ecommerce | Price/compare-at checks, image count, Shopify API | `maxtechera/skill-ecommerce` |\n| seo | Meta limits, word count, Lighthouse, GSC inspection | `maxtechera/skill-seo` |\n| engineering | Build/test/lint/CI gates, PR required, no self-merge | Built-in (agent default) |\n\nFull skill list and composition chains: [skills/README.md](skills/README.md)\n\n---\n\n## Key Concepts\n\n**Harness over trust** — the verification harness rules completion. No evidence, no done. This is the invariant.\n\n**The ticket contract** — every ticket has four sections: Inputs, Deliverables, Verification, Artifacts. Missing any → rejected to Backlog with guidance. [Full spec](docs/VISION.md).\n\n**Finishing beats starting** — tickets closest to done get dispatched first.\n\n**Self-improving harness** — every failure becomes a rule. Skills evolve, verification hardens, rules accumulate. All version-controlled. You approve every change.\n\n**Fail visible** — nothing ships without passing verification. Blocked tickets include a structured reason. Partial failures are caught. The system fails visibly, not silently.\n\n---\n\n## What This Repo Contains\n\n```\norchestrator/\n  SKILL.md                        # Core orchestrator skill — the agent reads this\n  WORKFLOW.md                     # Ticket lifecycle (intake → execute → verify → done)\n  .claude-plugin/                 # Claude Code marketplace manifests\n  .codex-plugin/                  # Codex CLI discovery\n  .agents/                        # OpenCode/OpenClaw skill discovery\n  gemini-extension.json           # Gemini CLI extension manifest\n  .clawhubignore                  # Distribution exclusions\n  .env.example                    # All configuration variables\n  CHANGELOG.md                    # Release history\n  docs/\n    VISION.md                     # Vision deck — the full story\n    STATE_MACHINE.md              # Ticket state transitions (detailed)\n  skills/                         # Domain skill directory\n  examples/                       # Example workflows and tickets\n```\n\n## Principles\n\n- Your ticket board is the single source of truth\n- The executor and the verifier are always different\n- `Done` is fail-closed: no evidence, no completion\n- The system fails visibly, not silently\n- Finishing beats starting\n- You define the boundaries. The system operates inside them.\n\n## Vision\n\nRead the full north star document: [`docs/VISION.md`](docs/VISION.md)\n\n**One board. Any domain. Verified output. The operator's leverage finally scales.**\n\n## License\n\nMIT\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmaxtechera%2Forchestrator","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fmaxtechera%2Forchestrator","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmaxtechera%2Forchestrator/lists"}