- Home
- Skills
- Agents & Orchestration
- Orchestra Board
Works with the AI Agents you already use
Orchestra Board
A tiered task orchestrator that routes coding sub-tasks to the cheapest viable Claude model with a live status board.
$9.99
Orchestra Board
Example session with this skill installed
Add a CSV export to the reports page, cover it with tests, and update the README. Use the orchestra board.
- Read your context and instructions
- Compiled the orchestra board
- Generated the data export
Here is the plan. The Orchestra board pane will show each worker live as it runs.
Plan
| # | Task | Tier | Acceptance check | Depends on |
|---|---|---|---|---|
| 1 | Add toCsv() in src/reports/export.ts and an Export CSV button on the reports page | standard (Sonnet) | npm test -- export passes, the button downloads report.csv | none |
| 2 | Escape commas, quotes and line breaks in CSV fields | light (Haiku) | the 6 cases in export.test.ts pass | 1 |
| 3 | Add an "Export to CSV" section to README.md | light (Haiku) | the section exists and npm run lint:md passes | 1 |
Execution
Task 1, Sonnet: SUCCESS. I re-ran npm test -- export myself: 4/4 passing.
Task 2, Haiku: FAILED.
I reverted export.ts to the end of task 1 and sent the task one tier up with the failure reason and "do not repeat this approach".
Task 2, Sonnet: SUCCESS. My re-run: 6/6 passing.
Task 3, Haiku: SUCCESS. The README section is added and npm run lint:md is clean.
Escalation log
| Task | Attempts (Haiku > Sonnet > Opus) | Final status | Reason for each failure |
|---|---|---|---|
| 1. CSV export | Sonnet | SUCCESS | none |
| 2. Field escaping | Haiku > Sonnet | SUCCESS | Haiku: quotes not doubled |
| 3. README | Haiku | SUCCESS | none |
Summary
- Files changed:
src/reports/export.ts,src/reports/ReportsPage.tsx,src/reports/export.test.ts,README.md - All checks verified by me, not just reported by the workers. No task needed Opus.
Next steps: try the button on a report that has many rows. Ask "what did my work cost this month?" for the 30-day token and cost report.
report.csv
CSV · data export
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
Stop burning expensive Opus tokens on boilerplate and formatting. Most coding tasks don't require the most powerful model, but manual switching is a friction point that slows you down.
This skill implements a hierarchical routing protocol that treats Claude as an orchestrator. It automatically offloads mechanical tasks to Haiku, standard features to Sonnet, and reserves Opus strictly for complex architectural logic or deep debugging.
What it does
- Tiered Routing automatically assigns tasks to Haiku, Sonnet, or Opus based on complexity.
- Failover Escalation detects task failures and automatically retries with a higher-tier model.
- Visual Task Tracking provides a side pane showing worker status, tool calls, and elapsed time.
- Cost Monitoring generates 30-day token and dollar estimates across all projects.
- Acceptance Verification mandates specific commands or checks to validate subagent output.
How it works
- Plan the request into a task table defining tiers (light, standard, deep) and acceptance checks.
- Delegate work to subagents via the Agent tool using model-specific tiers.
- Monitor progress via the live board plugin or the terminal status report.
- Escalate failed tasks by reverting files and retrying on the next model tier.
- Review the 30-day cost report by running the included Node.js utility.
Frameworks & tools
Built for Claude Code (2.1.316+). Requires Node.js 18+ for cost reporting. The visual board is optimized for the Claude desktop app but supports terminal fallback.
Installation
- Unzip the skill. Create the folder
~/.claude/skills/orchestra-board/and unziporchestra-board.zipinto it, so thatSKILL.mdsits directly inside (next to thepluginfolder). On Windows the folder isC:\Users\<you>\.claude\skills\orchestra-board\. Claude picks up the skill on its own. - Install the board. In a terminal, run:
On Windows, use the full path instead of
~, for example:claude plugin marketplace add "C:\Users\<you>\.claude\skills\orchestra-board\plugin" - Reload. Restart Claude Code, or run
/reload-pluginsin an open session. The board opens at the start of every session, and/boardreopens it. - Optional: to apply the protocol to every coding task without asking, add this line to
~/.claude/CLAUDE.md:For multi-step coding tasks, follow the orchestra-board skill.
To turn the board off: claude plugin disable orchestra-board@orchestra. To remove it: claude plugin uninstall orchestra-board@orchestra, then delete the skill folder.
Why this beats prompting it yourself
Manual model switching is tedious and often forgotten, leading to wasted spend on simple edits. This skill bakes the "start cheap, escalate on failure" logic into the agent's core loop, ensuring consistent cost-efficiency without manual intervention.
Use cases
- Large Refactors where formatting and renames go to Haiku while logic stays on Sonnet.
- Cost Auditing to see exactly how many tokens your projects consumed over the last month.
- Parallel Feature Dev by running independent sub-tasks across different model tiers simultaneously.
Known limitations
The board plugin needs a one-time install from the terminal (two commands, see Installation). Cost reporting relies on list-price assumptions and transcript file availability.
Requires Claude Code 2.1.316+ and Node.js 18+. Cost reports are estimates based on list prices. Destructive tasks require manual confirmation before escalation.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed today
Requires